Method and apparatus for detecting a desired signal in the presence of an interfering signal

Abstract

In accordance with an embodiment, there is provided a method comprising receiving, at a receiver, a desired signal and an interfering signal, wherein the interfering signal was transmitted with a modulation unknown to the receiver; identifying a likely modulation corresponding to the modulation with which the interfering signal was transmitted; and decoding the desired signal using a modulation dependent multiple-input multiple output (MIMO) detection, wherein the modulation dependent MIMO detection is based at least in part on the identified likely modulation corresponding to the modulation with which the interfering signal was transmitted, wherein the modulation dependent MIMO detection includes maximum likelihood (ML) detection.

Description

CROSS-REFERENCE TO RELATED APPLICATION

The present disclosure claims priority to U.S. Provisional Patent Application No. 61/551,339, filed on Oct. 25, 2011, which is incorporated herein by reference.

TECHNICAL FIELD

This disclosure relates to multiple-input multiple-output (MIMO) systems, and more specifically, to receivers implemented in MIMO systems.

BACKGROUND

The background description provided herein is for the purpose of generally presenting the context of the disclosure. Work of the presently named inventors, to the extent it is described in this background section, as well as aspects of the description that may not otherwise qualify as prior art at the time of filing, are neither expressly nor impliedly admitted as prior art against the present disclosure.

In wireless communication systems, multiple-input and multiple-output (MIMO) refers to an use of multiple antennas at both a transmitter and a receiver to improve communication performance. MIMO technology has attracted attention in wireless communications, because MIMO offers significant increase in data throughput and link range without additional bandwidth or increased transmit power. MIMO achieves this by, for example, spreading the same total transmit power over the antennas to achieve an array gain that improves the spectral efficiency (e.g., more bits per second per hertz of bandwidth) or to achieve a diversity gain that improves a link reliability (e.g., reduced fading). Because of these properties, MIMO is an important part of various modern wireless communication systems.

SUMMARY

In accordance with an embodiment, there is provided a method comprising receiving, at a receiver, a desired signal and an interfering signal, wherein the interfering signal was transmitted with a modulation unknown to the receiver; identifying a likely modulation corresponding to the modulation with which the interfering signal was transmitted; and decoding the desired signal using a modulation dependent multiple-input multiple output (MIMO) detection, wherein the modulation dependent MIMO detection is based at least in part on the identified likely modulation corresponding to the modulation with which the interfering signal was transmitted, wherein the modulation dependent MIMO detection includes maximum likelihood (ML) detection.

There is also provided, in accordance with an embodiment, a system comprising one or more antennas configured to receive a desired signal and an interfering signal, wherein the interfering signal was transmitted with a modulation unknown to the system; and a data processing unit coupled to the one or more antennas, wherein the data processing unit is configured to (i) identify a likely modulation or the likelihood of a modulation corresponding to the modulation with which the interfering signal was transmitted, and (ii) decode the desired signal using a modulation dependent detection, wherein the modulation dependent detection is based at least in part on the identified likely modulation or the identified likelihood of the modulation corresponding to the modulation with which the interfering signal was transmitted, wherein the modulation dependent detection includes maximum likelihood (ML) detection.

BRIEF DESCRIPTION OF THE DRAWINGS

Embodiments of the present disclosure will be readily understood by the following detailed description in conjunction with the accompanying drawings. To facilitate this description, like reference numerals designate like elements.

FIG. 1 is a block diagram illustrating a communications system.

FIG. 2 is a flowchart illustrating a method of decoding a desired signal using maximum likelihood detection.

FIG. 3 is a block diagram illustrating a method of decoding a desired signal using interference cancellation.

DETAILED DESCRIPTION

Described herein are example systems, components, and methods that may be used in conjunction with MIMO (multiple-input/multiple-output) and/or MU-MIMO (multi-user MIMO) techniques. The following description merely provides examples and is in no way intended to limit the disclosure, its application, or uses.

A MIMO receiver (also referred to herein more generally as a “receiver”) typically receives a signal that carries a plurality of effective data signals, only some of which is intended for the receiver. The effective data signal intended for the receiver, referred to herein as the desired signal, may be decoded using a corresponding channel transfer matrix, also referred to as a channel matrix. Other effective data signals, referred to as interfering signals, are associated with different channel matrices. In some cases, the channel matrices of interfering signals may be known or may be estimated.

The effective signals described above may be transmitted using different modulations. Available modulations may include, for example, various implementations of quadrature amplitude modulation (QAM), quadrature phase shift keying (QPSK), and others. In some situations, the modulations used by interfering signals may not be known to a receiver. This may be the case, for example, when interfering signals are the result of multi-user MIMO or co-channel interference.

This disclosure is directed to mitigating interference in an MIMO environment in which a receiver is able to identify presence of interfering signals and their corresponding channel matrices, but in which the modulations of the interfering signals are not known to the receiver.

FIG. 1 illustrates an example multiple-user multiple-input/multiple-output (MU-MIMO) communication system 100. Three mobile devices (e.g., mobile devices that communicate data by way of radio transmissions) U₁, U₂, and U₃ are interfacing with the system 100. The system 100 includes two base stations (e.g., cellular telephone transmitter/receivers towers) BS₁ and BS₂. An approximate boundary for each base station's range, referred to as the base station's “cell”, is shown as a circle surrounding the base station. While the systems and methods herein will be described in the context of a multiuser MIMO environment, the systems and method may be used in any communication system in which a signal detection technique accounts for the potential existence of a number of interfering communication channels.

Base station BS₁ communicates with the mobile device U₁ by way of an effective signal, which is defined or characterized by a channel matrix h_1,1. BS₁ also communicates with the mobile device U₂ by way of effective signal defined or characterized by a channel matrix h_2,2. The signal being transmitted to U₂ may interfere with the signal being transmitted to U₁. This intra-cell interfering signal travels along an effective channel characterized by a channel matrix h_1,2 to U₁.

Because many modern wireless communication systems have adjacent cells transmitting in the same frequency bands, mobile devices are also subject to inter-cell interference. The base station BS₂ exchanges signals with the mobile device U₃ by way of an effective channel characterized by the channel matrix h_3,3. This signal also interferes with the signal being transmitted by BS₁ to U₁.

The following discussion focuses on operations performed at a receiver that receives multiple signals, each of which is characterized by a different channel matrix h_i, where the suffix is an index corresponding to an individual signal and channel matrix. Thus, if a receiver receives two effective signals x₁ and x₂, those effective signals may be characterized by channel matrices h₁ and h₂, respectively. While the following discussion focuses on the two data streams, the teachings of this disclosure applies to systems that can handle a larger number of data streams.

One embodiment of a wireless signal processing unit 102 is shown in FIG. 1. The wireless signal processing unit 102 is part of the mobile device U₁. The wireless signal processing unit 102 includes a MIMO transmitter/receiver radio 104. The radio 104 includes four antennas 106 that are configured to receive a signal r containing multiple effective signals x. The radio 104 provides the signal r to a baseband processor 108. The baseband processor 108 combines performing signal detection and/or filtering based on information about the channels of the received signal to estimate a desired signal meant for the particular mobile device U₁.

The signal processing unit 102 from FIG. 1 may be implemented on a chip including one or more integrated circuits configured to perform one or more of the functions described herein. The signal processing unit 102 may be implemented in a computing device, for example, a computer, a laptop, a server, a cell phone, a hand held computing device, or other type of device that uses memory. The baseband processor 108 may be part of the same device as the signal processing unit 102 or may be external to the apparatus. The signal processing unit 102 may be configured by way of instructions stored in associated computer-readable media, which are executable by the processing unit 102 to perform the computations, calculations, and/or operations described below.

The general characteristics of MIMO transmission, assuming a number of users K, may be characterized as follows:

$r\left(i\right)=H\left(i\right)x\left(i\right)+n\left(i\right)=H_1\left(i\right)x_1\left(i\right)+\sum _{k=2}^K{H}_k\left(i\right)x_k\left(i\right)+n\left(i\right)1\le i\le L$

r: N_R×1 received vector

$H\text{:}{N}_R\times \sum _{k=1}^K{N}_{T,k}$ effective MIMO channel, where H=[H₁ . . . H_K]

$x\text{:}\sum _{k=1}^K{N}_{T,k}\times 1$ transmitted vector

x₁: transmitted signal for the desired user of modulation M⁽¹⁾

x_k: transmitted signal for the interferer of modulation M^(k), k=2 . . . K

n: white noise at receiver, such that E(nn^H)=σ²I

N_R: number of receiver antennas

N_T,k: number of transmit antennas (number of spatial streams) of each user

i: a transmission instance out of total L transmissions, such as an index of subcarriers.

Equation 1

For ease of explanation, the following discussion will assume two rank-1 users, resulting in the following simplification of Equation 1:r(i)=H(i)x(i)+n(i)=h₁(i)x₁(i)+h₂(i)x₂(i)+n(i)1≦i≦L

• • H: N_R×2 effective MIMO channel, where H=[h₁ h₂]
  • x: 2×1 transmitted vector
  • x₁: transmitted signal for the desired user of modulation M⁽¹⁾

x₂: transmitted signal for the interferer of modulation M⁽²⁾, M⁽²⁾εΨ

Equation 2

The channel or channel matrix associated with the desired signal x₁ is represented as h₁. The channel or channel matrix associated with the interfering signal x₂ is represented as h₂.

It is assumed in the following discussion that the existence of the interfering signal is known, and that the channel matrix h₂ of the interfering channel has been obtained from advanced channel estimation techniques. For example, the channel matrix h₂ may be calculated as the orthogonal RS sequence of transmission mode 8 of the “Long Term Evolution” (LTE) Rel. 9 specification.

It is further assumed that the modulation M⁽²⁾ of the interfering signal x₂ is chosen from a finite set ψ of possible modulations, and that the composition of this set is known by the receiver. For example, if the receiver knows the wireless system from which the interference comes from, the receiver may also know a set of all possible modulations for the interference. However, the particular modulation M⁽²⁾ being used by the interfering signal x₂ is not known at the receiver.

The interfering effective signal may be associated with a co-scheduled MU-MIMO channel, or may be associated with a co-channel in a neighboring system (e.g., a neighboring cell for a cellular system, or a nearby access point for a WiFi system).

Generally, an MMSE receiver can be applied to detect a received signal x₁, without knowledge of the modulation of an interfering signal x₂, as follows:{tilde over (x)}₁=[H^H(HH^H+σ²I)⁻¹]_(1,:)r=h₁^H(HH^H+σ²I)⁻¹r{circumflex over (x)}₁=Q({tilde over (x)}₁;M⁽¹⁾)

Equation 3

The MMSE receiver of Equation 3 works regardless of the modulation used in the interfering signal, and is a conventional solution for interference mitigation. In Equation 3, Q(.;M) is a slicer based on the modulation M, as an example for decision. In some embodiments, a hard slicer is replaced by a soft decision (e.g., a bit-wise log-likelihood-ratio, or LLR) calculator taking the MMSE receiver output {tilde over (x)}₁ as input.

In cases where the modulation M⁽²⁾ of an interfering signal is known, a modulation-dependent receiver may outperform a modulation independent receiver such as the MMSE receiver, in which the interference is treated as Gaussian noise. For example, ML detection as an example of modulation-dependent receiver can be applied as follows:

$\begin{array}{cc}{\hat{x}}_1=\underset{x_1\in {\Omega }^{\left(M^{\left(1\right)}\right)},x_2\in {\Omega }^{\left(M^{\left(2\right)}\right)}}{\arg \min }{r-h_1{x}_1-h_2{x}_2}^2& \mathrm{Equation}4\end{array}$ Equation 4 shows an example of hard decision using ML criterion. A soft decision (such as bit-wise LLR) can also be used in some embodiments following ML criterion. Other examples of modulation dependent receiver include decision-feedback equalizer, interference cancellation, and/or the like. In some of the embodiments discussed herein later, ML detection is used as an example.

However, the modulation M⁽²⁾ of the interfering effective signal x₂ is often unknown. Described below are various methods, techniques, and enhancements for performing ML detection even in situations where the modulation M⁽²⁾ of the interfering signal is unknown.

FIG. 2 illustrates an example of decoding a desired signal x₁ associated with a receiver channel h₁, in the presence an interfering signal x₂ associated with an interfering channel h₂. Although the example shows a single interfering signal, the described techniques may be used in a similar fashion to decode the desired signal x₁ in the presence of multiple interfering channels.

An action 202 comprises receiving the desired signal x₁ and the interfering signal x₂. The signals may be generated by different transmitters or base stations, and/or from different antennas of a single transmitter or base station. The modulation of the interfering signal x₂ is unknown to the receiver.

An action 204 comprises identifying a likely modulation of the interfering signal x₂. This may be performed in a number of different ways, as will be described below. In certain embodiments, the action 204 may comprise estimating, calculating, or otherwise determining a modulation {circumflex over (M)}⁽²⁾ that is assumed to be used by the interfering channel h₂.

The action 204 may be generally characterized by the following equation:{circumflex over (M)}⁽²⁾=f(h₁,h₂,r,M⁽¹⁾,σ²)

Equation 5

An action 206 comprises applying maximum likelihood (ML) detection, based at least in part on the assumed modulation {circumflex over (M)}⁽²⁾, in order to obtain the desired signal x₁. An example of ML detection to determine a hard estimate of x₁ is generally characterized by the following equation:

$\begin{array}{cc}{\hat{x}}_1=\underset{x_1\in {\Omega }^{\left(M^{\left(1\right)}\right)},x_2\in {\Omega }^{\left(M^{\left(2\right)}\right)}}{\arg \min }{r-h_1{x}_1-h_2{x}_2}^2& \mathrm{Equation}6\end{array}$ where Ω is the set of possible or available modulations that may have been used by the desired signal and the interfering signal. A bit-wise soft decision can also be determined in some embodiments in a similar fashion with the assumed modulation {circumflex over (M)}⁽²⁾. In the following discussion and as an example, hard decision is used without excluding the application of the techniques to soft decision.

The actions 204 and 206 may in some cases be combined, using calculations that jointly determine both the desired signal x₁ and the assumed modulation {circumflex over (M)}⁽²⁾ as follows:{{circumflex over (x)}₁,{circumflex over (M)}⁽²⁾}=G(h₁,h₂,r,M⁽¹⁾,σ²)

Equation 7

The calculations of Equations 5, 6, and 7 are implemented to minimize a cost function over all possible modulations:

$\begin{array}{cc}{\hat{M}}^{\left(2\right)}=\underset{M^{\left(2\right)}\in \Psi }{\min }{f}_{M^{\left(2\right)}}\left(h_1,h_2,r,M^{\left(1\right)},{\sigma }^2\right)& \mathrm{Equation}8\end{array}$ Joint ML Sequence and Modulation Detection

An example of applying Equation 7 is that the actions 204 and 206 may be performed jointly, by evaluating the a-posteriori probabilities of the different available modulations of the set ψ in combination with the decoded desired signal x₁. Evaluation of a-posteriori probabilities may be performed in several different ways, as indicated along the right side of FIG. 2, under the heading “Joint”.

Log-MAP

A method referred to herein as “Log-MAP” may be used to maximize the a-posteriori probability of the desired signal x₁ and the assumed modulation {circumflex over (M)}⁽²⁾ of the interfering signal x₂, given the signal r, as follows:

$\begin{array}{cc}\left\{{\hat{x}}_1\left(1\right)\dots {\hat{x}}_1\left(L\right),{\hat{M}}^{\left(2\right)}\right\}=\underset{M^{\left(2\right)}\in \Psi }{\arg \max }P\left(x_1\left(1\right)\dots x_1\left(L\right),M^{\left(2\right)}|r\right)& \mathrm{Equation}9\end{array}$ Where L is the number of signals in the sequence.

Eq. 9 may be implemented, in an embodiment, as follows:

$\mathrm{Equation}10$ $\left\{{\hat{x}}_1\left(1\right)\dots {\hat{x}}_1\left(L\right),{\hat{M}}^{\left(2\right)}\right\}={\underset{M^{\left(2\right)},x_1,x_2}{\mathrm{argmax}}\left(\frac{1}{M^{\left(2\right)}}\right)}^L\prod _{i=1}^L\left(\sum _{m=1}^{M^{\left(1\right)}}\sum _{n=1}^{M^{\left(2\right)}}\exp \left(-\frac{{r\left(i\right)-h_1\left(i\right)x_1^{\left(m\right)}-h_2\left(i\right)x_2^{\left(n\right)}}^2}{{\sigma }^2}\right)\right)$

Equation 10 assumes that any residual noise is Gaussian, and that each of the available modulations has an equal a-priori probability of being used by the interfering channel. Equation 10 finds the a-posteriori probabilities of all possible modulations, and then selects the modulation having the highest a-posteriori probability.

In some situations, certain modulations are more likely to be used than others. In these situations, Eq. 9 may be implemented as follows, to account for the differing a-priori probabilities of each modulation:

$\mathrm{Equation}11$ $\left\{{\hat{x}}_1\left(1\right)\dots {\hat{x}}_1\left(L\right),{\hat{M}}^{\left(2\right)}\right\}=\underset{M^{\left(2\right)},x_1,x_2}{\mathrm{argmax}}P\left(M^{\left(2\right)}\right){\left(\frac{1}{M^{\left(2\right)}}\right)}^L\prod _{i=1}^L\left(\sum _{m=1}^{M^{\left(1\right)}}\sum _{n=1}^{M^{\left(2\right)}}\exp \left(-\frac{\begin{array}{c}r\left(i\right)-h_1\left(i\right)x_1^{\left(m\right)}-\\ {h_2\left(i\right)x_2^{\left(n\right)}}^2\end{array}}{{\sigma }^2}\right)\right)$ In the following discussion and as an example, the equal a-priori probabilities of each modulation is used without excluding the scenarios of unequal a-priori probabilities of each modulation.Sum-Max-Log-MAP

The Log-MAP method of maximizing the a-posteriori probability of the desired signal x₁ and the modulation {circumflex over (M)}⁽²⁾ of the interfering channel may be refined using a method referred to herein as “Sum-Max-Log-MAP”. Specifically, equations 10 and 11 can be implemented in a less complex fashion by approximating the sum of the a-posteriori probabilities, where a core part of Equations 10 and 11 is replaced by a simplification. In particular, the term

$\frac{1}{M^{\left(2\right)}}\sum _{m=1}^{M^{\left(1\right)}}\sum _{n=1}^{M^{\left(2\right)}}\exp \left(-\frac{{r\left(i\right)-h_1\left(i\right)x_1^{\left(m\right)}-h_2\left(i\right)x_2^{\left(n\right)}}^2}{{\sigma }^2}\right)$ of Equations 10 and 11 may be replaced by the following expression:

$\begin{array}{cc}\frac{1}{M^{\left(2\right)}}\exp \left(-\frac{\min {r\left(i\right)-h_1\left(i\right)x_1^{\left(m\right)}-h_2\left(i\right)x_2^{\left(n\right)}}^2}{{\sigma }^2}\right)& \mathrm{Equation}12\end{array}$ The Approximation Leads to Little Performance Loss, for Example, when Signal to Noise Ratio (SNR) is Relatively High.

For example, when the probability of each M2 is equal, the desired signal x₁ and the modulation {circumflex over (M)}⁽²⁾ of the interfering signal x₂ are then obtained by comparing the average minimum distance per modulation with an offset:

$\mathrm{Equation}13$ $\begin{array}{c}\left\{{\hat{x}}_1\left(1\right)\dots {\hat{x}}_1\left(L\right),{\hat{M}}^{\left(2\right)}\right\}={\underset{M^{\left(2\right)},x_1{x}_2}{\mathrm{argmax}}\left(\frac{1}{M^{\left(2\right)}}\right)}^L\exp \left(-\sum _{l=1}^L\frac{\begin{array}{c}\min r\left(i\right)-h_1\left(i\right)x_1^{\left(m\right)}-\\ {h_2\left(i\right)x_2^{\left(n\right)}}^2\end{array}}{{\sigma }^2}\right)\\ =\mathrm{argmin}\left(\frac{1}{L}\sum _{i=1}^L\frac{\begin{array}{c}\min r\left(i\right)-h_1\left(i\right)x_1^{\left(m\right)}-\\ {h_2\left(i\right)x_2^{\left(n\right)}}^2\end{array}}{{\sigma }^2}+\log \left(M^{\left(2\right)}\right)\right)\end{array}$ Mean-Max-Log-MAP

Alternatively, for example, when the SNR is relatively low, Equations 10 and 11 can be implemented using a method referred to herein as “Mean-Max-Log-MAP”, which comprises approximating the mean probability by the maximum probability. Specifically, the term

$\frac{1}{M^{\left(2\right)}}\sum _{m=1}^{M^{\left(1\right)}}\sum _{n=1}^{M^{\left(2\right)}}\exp \left(-\frac{{r\left(i\right)-h_1\left(i\right)x_1^{\left(m\right)}-h_2\left(i\right)x_2^{\left(n\right)}}^2}{{\sigma }^2}\right)$ of Equations 10 and 11 is replaced by the following expression:

$\begin{array}{cc}\exp \left(-\frac{\min {r\left(i\right)-h_1\left(i\right)x_1^{\left(m\right)}-h_2\left(i\right)x_2^{\left(n\right)}}^2}{{\sigma }^2}\right)& \mathrm{Equation}14\end{array}$

For example, when the probability of each M2 is equal, the desired signal x₁ and modulation {circumflex over (M)}⁽²⁾ of the interfering signal are then obtained as follows:

$\begin{array}{cc}\left\{{\hat{x}}_1\left(1\right)\dots {\hat{x}}_1\left(L\right),{\hat{M}}^{\left(2\right)}\right\}=\underset{M^{\left(2\right)},x_1,x_2}{\arg \min }\left(\frac{1}{L}\sum _{i=1}^L\frac{\min {r\left(i\right)-h_1\left(i\right)x_1^{\left(m\right)}-h_2\left(i\right)x_2^{\left(n\right)}}^2}{{\sigma }^2}\right)& \mathrm{Equation}15\end{array}$ Max-Log-MAP-Scale

In cases where it is difficult or impractical to make a decision regarding SNR, the Equations 10 and/or 11 may be implemented as follows, independently of SNR values:

$\begin{array}{cc}\begin{array}{c}\left\{{\hat{x}}_1\left(1\right)\dots {\hat{x}}_1\left(L\right),{\hat{M}}^{\left(2\right)}\right\}=\underset{M^{\left(2\right)},x_1,x_2}{\arg \max }{\left(\frac{1}{M^{\left(2\right)}}\right)}^L\prod _{i=1}^L(\frac{1+\alpha \left(H,M^{\left(1\right)},M^{\left(2\right)},{\sigma }^2\right)}{M^{\left(2\right)}}·\\ \exp \left(-\frac{\min {r\left(i\right)-h_1\left(i\right)x_1^{\left(m\right)}-h_2\left(i\right)x_2^{\left(n\right)}}^2}{{\sigma }^2}\right))\\ =\underset{M^{\left(2\right)},x_1,x_2}{\arg \max }\frac{1}{L}\sum _{i=1}^L(\frac{\min {r\left(i\right)-h_1\left(i\right)x_1^{\left(m\right)}-h_2\left(i\right)x_2^{\left(n\right)}}^2}{{\sigma }^2}-\\ \log \left(1+\alpha \left(H,M^{\left(1\right)},M^{\left(2\right)},{\sigma }^2\right)\right))+\log \left(M^{\left(2\right)}\right)\end{array}& \mathrm{Equation}16\end{array}$

In Equation 16, the term

$\frac{1+\alpha \left(H,M^{\left(1\right)},M^{\left(2\right)},{\sigma }^2\right)}{M^{\left(2\right)}}$ is a scaling factor that is used to scale the subsequent exponential term of Equation 16. The scaling factor can be approximated and stored as a lookup table to lessen the complexity of calculating Equation 16. The desired signal x₁ and the modulation {circumflex over (M)}⁽²⁾ of the interfering channel x₂ may be calculated in this scenario as follows:

$\begin{array}{cc}\left\{{\hat{x}}_1\left(1\right)\dots {\hat{x}}_1\left(L\right),{\hat{M}}^{\left(2\right)}\right\}=\underset{M^{\left(2\right)},x_1,x_2}{\arg \min }\frac{1}{L}\sum _{i=1}^L\left(\frac{\min {r\left(i\right)-h_1\left(i\right)x_1^{\left(m\right)}-h_2\left(i\right)x_2^{\left(n\right)}}^2}{{\sigma }^2}\right)+I\left(r;x_1,x_2|M^{\left(1\right)},M^{\left(2\right)},H\right)& \mathrm{Equation}17\end{array}$ Log-MAP-Best-K

As yet another example, a balance between accuracy and complexity may be obtained when calculating Equations 10 and 11 using a method referred to herein as “log-MAP-Best-K”, which comprises computing only the largest K terms in the sum of conditional probabilities, as follows:

$\begin{array}{cc}\left\{{\hat{x}}_1\left(1\right)\dots {\hat{x}}_1\left(L\right),{\hat{M}}^{\left(2\right)}\right\}\approx \hspace{1em}\underset{M^{\left(2\right)},x_1,x_2}{\arg \max }\frac{1}{M^{\left(2\right)}}\sum _{k=1}^{K\left(M^{\left(2\right)}\right)}\exp (-\frac{{r\left(i\right)-h_1\left(i\right)x_1^{\left(k,m\right)}-h_2\left(i\right)x_2^{\left(k,n\right)}}^2}{{\sigma }^2})s.t.{r\left(i\right)-h_1\left(i\right)x_1^{\left(1,m\right)}-h_2\left(i\right)x_2^{\left(1,n\right)}}^2\le {r\left(i\right)-h_1\left(i\right)x_1^{\left(2,m\right)}-h_2\left(i\right)x_2^{\left(2,n\right)}}^2\le \dots {r\left(i\right)-h_1\left(i\right)x_1^{\left(M^{\left(1\right)},M^{\left(2\right)},m\right)}-h_2\left(i\right)x_2^{\left(M^{\left(1\right)},M^{\left(2\right)},n\right)}}^2& \mathrm{Equation}18\end{array}$ The depth of this calculation may be dependent on one or more factors including modulation, SNR, and/or the like, and may be adjusted in the real-time.Sequential Modulation and Sequence Detection

In certain embodiments, the assumed modulation M⁽²⁾ may be initially determined, and may then be used as the basis for decoding the desired signal x₁. This is referred to herein as sequential modulation and sequence detection. The assumed modulation M⁽²⁾ may be calculated or estimated in various ways, as indicated on the left side of FIG. 2 under the heading “Sequential”.

The techniques for joint ML sequence and modulation detection mentioned above can also be implemented in the fashion of sequential modulation and sequence detection discussed hereafter. For example, the modulation of M2 can be detected by maximizing the a-posteriori probability of the desired signal x₁ and the assumed modulation {circumflex over (M)}⁽²⁾ of the interfering signal x₂, given the signal r, as follows:

$\begin{array}{cc}{\hat{M}}_2=\underset{M_2\in \Psi }{\arg \max }P\left(x_1\left(1\right)\dots x_1\left(L\right),M_2|r\right)& \mathrm{Equation}19\end{array}$ And then the determined M2 is used for a ML detection of the desired sequence. That is,

${\hat{x}}_1=\underset{x_1\in {\Omega }^{\left(M^{\left(1\right)}\right)},x_2\in {\Omega }^{\left(K\right)}}{\arg \min }{r-h_1{x}_1-h_2{x}_2}^2$ The equations [9] to [18] can be similarly implemented in the sequential fashion. Additionally, some specific examples for sequential modulation and sequence detection are discussed as follows.Fixed Modulation

As an example of sequentially determining M⁽²⁾ and x₁, the action 204 may be performed by arbitrarily or randomly assuming a modulation {circumflex over (M)}⁽²⁾ of the interfering signal. In this case, {circumflex over (M)}⁽²⁾=K, where K is an arbitrarily assumed modulation belonging to the known set ψ of available modulations. The desired signal x₁ may be calculated in the action 206 by performing ML detection based on this assumption of a fixed modulation for the interfering signal, as follows:

$\begin{array}{cc}{\hat{x}}_1=\underset{x_1\in {\Omega }^{\left(M^{\left(1\right)}\right)},x_2\in {\Omega }^{\left(M^{\left(2\right)}\right)}}{\arg \min }{r-h_1{x}_1-h_2{x}_2}^2& \mathrm{Equation}20\end{array}$

As another example of sequentially determining M⁽²⁾ and x₁, the action 204 may be performed by assuming the superposition of all possible modulations of the interfering signal. Thus, assuming that the known set ψ of modulations comprises, for example, QPSK, QAM16, and QAM64 modulations, the action 206 of calculating the desired signal x₁ can be calculated by performing ML detection as follows:

$\begin{array}{cc}{\hat{x}}_1=\underset{x_1\in {\Omega }^{\left(M^{\left(1\right)}\right)};x_2\in {\Omega }^{\left(\mathrm{QPSK}\right)}\bigcup {\Omega }^{\left(\mathrm{QAM}16\right)}\bigcup {\Omega }^{\left(\mathrm{QAM}64\right)}}{\arg \min }{r-h_1{x}_1-h_2{x}_2}^2& \mathrm{Equation}21\end{array}$ Linear Receiver with Modulation Detection

Sequential determination of M⁽²⁾ and x₁ may also be performed by implementing a linear receiver for the interfering signal x₂, to detect the modulation {circumflex over (M)}⁽²⁾ of x₂. The action 204 may be performed by a linear receiver using MMSE, zero-forcing (ZF), maximal ratio combining (MRC), and/or other types of linear detection schemes. In an embodiment, an MMSE receiver may implemented as follows:

$\begin{array}{cc}{\tilde{x}}_2=C_{\mathrm{tmear}}r\stackrel{\mathrm{MMSE}}{=}\alpha {h_2^H\left({\mathrm{HH}}^H+{\sigma }^{2}I\right)}^{-1}r=x_2+z{\sigma }_z^2=E\left[{z}^2\right]=\frac{1}{{\mathrm{SINR}}_{\mathrm{MMSE}}}& \mathrm{Equation}22\end{array}$

Based on Equation 22, a modulation detector may be implemented to determine an assumed modulation {circumflex over (M)}⁽²⁾ of the interfering signal x₂ as follows:

$\begin{array}{cc}{\hat{M}}^{\left(2\right)}=\underset{M^{\left(2\right)}\in \Psi }{\arg \min }J\left({\hat{x}}_2\left(1\right)\dots {\tilde{x}}_2\left(L\right),M^{\left(2\right)},{\sigma }_z^2\right)& \mathrm{Equation}23\end{array}$ The optimization function J(.) can be derived from the counterpart equations [9] to [18]. The analogy between the optimization functions is that h₁ and x₁ equal to 0, and h₂ equal to 1, also r equal to {tilde over (x)}₂. In other word, the relationship between the transmitted and received signals in Equation [1] and [2] is reduced to a single user case, where the received signals are the output of the linear receiver {tilde over (x)}₂ and the transmitted signal x₂ travels through an AWGN channel in absence of x₁. By applying the principle of modulation detection, or a-posteriori probability determination in Equation [9] to [18] for the new input-output relationship, the modulation of x₂ may be determined accordingly. For example, the log-MAP style function is as follows:

${\hat{M}}_2=\underset{M_2\in \psi }{\arg \max }{\left(\frac{1}{M^{\left(2\right)}}\right)}^L\prod _{i=1}^L\left(\sum _{n=1}^{M^{\left(2\right)}}\exp \left(-\frac{{{\tilde{x}}_2\left(i\right)-x_2^{\left(n\right)}}^2}{{\sigma }_z^2}\right)\right)$ The technique of sum-max-log-MAP, mean-max-log-MAP, max-log-MAP-scale, log-MAP-best-K can be applied similarly.

After determining the assumed modulation of the interfering signal in this manner, the desired signal x₁ may be calculated in the action 206 by using ML detection as follows:

$\begin{array}{cc}{\hat{x}}_1=\underset{x_1\in {\Omega }^{\left(M^{\left(1\right)}\right)},x_2\in {\Omega }^{\left({\hat{M}}^{\left(2\right)}\right)}}{\arg \min }{r-h_1{x}_1-h_2{x}_2}^2& \mathrm{Equation}24\end{array}$ LLR-Based Modulation Detection

The modulation of the interfering signal may also be determined or estimated in the action 204 based on a soft decision or log-likelihood ratio (LLR) of each bit at the output of a receiver, as follows:

$\begin{array}{cc}{\mathrm{LLR}}_k=\log \left(\frac{P\left(b_1\left(k\right)=1|r\right)}{P\left(b_1\left(k\right)=0|r\right)}\right)& \mathrm{Equation}25\end{array}$

Instead of examining the probability of each symbol in a transmitted sequence, this technique evaluates the probability of each coded bit:

$\begin{array}{cc}{\mathrm{LLR}}_k^{\left(M^{\left(2\right)}\right)}=\log \left(\frac{P\left(b_1\left(k\right)=1|r,M^{\left(2\right)}\right)}{P\left(b_1\left(k\right)=0|r,M^{\left(2\right)}\right)}\right){\hat{M}}^{\left(2\right)}=\underset{M_2\in \Psi }{\arg \max }J\left({\mathrm{LLR}}_k,M^{\left(2\right)}\right)& \mathrm{Equation}26\end{array}$

For example, the following metric can represent the probability of a correctly decoded bit sequence:

$\begin{array}{cc}J\left(M^{\left(2\right)}\right)=\prod _{k=1}^{L^{\prime }}\frac{1}{1+\exp \left(-{\mathrm{LLR}}_k^{\left(M^{\left(2\right)}\right)}\right)}& \mathrm{Equation}27\end{array}$

Alternatively, an approximate and simpler solution is as follows:

$\begin{array}{cc}J\left(M^{\left(2\right)}\right)=\prod _{k=1}^{L^{\prime }}{\mathrm{LLR}}_k^{\left(M^{\left(2\right)}\right)}& \mathrm{Equation}28\end{array}$ Joint Decoding and Modulation Detection

In cases where signals x comprise coded symbols, modulation detection and decoding can be performed jointly or concurrently to determine the most probable modulation {circumflex over (M)}⁽²⁾ of the interfering signal x₂. For example, multiple ML detection may be performed in parallel or in sequence, based on all possible M⁽²⁾ modulations. The resulting outputs may then be fed to channel decoders to produce decoded sequences. Assuming that the symbols of the sequences have error encoding, such as cyclical redundancy check (CRC) encoding, the decoded sequences may be examined to determine whether they exhibit errors. The output sequence exhibiting the lowest error rate or the highest accuracy may then be selected, and the modulation M resulting in that output sequence may be assumed to be the modulation M⁽²⁾ of the interfering signal.

When CRC is not available, other approaches may be used to select the output sequence that is most likely to be correctly decoded. For example, the mean of the absolute information bit LLR may be used to determine whether an output stream has been correctly decoded.

After determining M⁽²⁾ in this manner, ML detection may be applied in the action 206 to determine the desired signal x₁ based on the determined modulation M⁽²⁾ of the interfering signal.

Hard Decision Distribution Metric

In an embodiment, the constellation distribution of hard decisions in an MMSE receiver, applied to the interfering channel x₂, may be evaluated to determine the likely validity of the previously described decisions regarding M⁽²⁾. For example, the hard decisions regarding the interfering channel x₂ may be evaluated to determine whether they exhibit expected constellation distributions for the assumed modulation {circumflex over (M)}⁽²⁾.

For example, a hard slicer may be applied at the output of a linear receiver such as an MMSE receiver, as follows:{circumflex over (x)}₂=Q(αh₂^H(HH^H+σ²I)⁻¹r;M⁽²⁾)

Equation 29

A typical transmitted signal uniformly occupies the constellation. If the hard decision at the output of hard slicer shows a non-uniform distribution in the assumed constellation, the likelihood of this assumed modulation, P(M⁽²⁾), may decrease, or this assumption may become invalid.

For example, for M⁽²⁾>4, the numbers of hard decisions in different subsets of the constellation of the assumed modulation {circumflex over (M)}⁽²⁾ are calculated, and these numbers are compared to expected distributions within the assumed modulation {circumflex over (M)}⁽²⁾. As an example, the constellation of the assumed modulation {circumflex over (M)}⁽²⁾ may be partitioned into two subsets, and the following ratio may be calculated:

$\begin{array}{cc}{\rho }^{\left(M_2\right)}=\frac{\left\{{\hat{x}}_2\left(i\right):{\hat{x}}_2\left(i\in {\Omega }_1⋐{\Omega }^{\left(M^{\left(2\right)}\right)}\right)\right\}}{\left\{{\hat{x}}_2\left(i\right):{\hat{x}}_2\left(i\in {\Omega }_2⋐{\Omega }^{\left(M^{\left(2\right)}\right)}\right)\right\}}& \mathrm{Equation}30\end{array}$

In Equation 30, subset 1 (Ω₁) includes constellation points of the assumed modulation {circumflex over (M)}⁽²⁾ that are relatively likely if the assumed modulation is actually used by the interfering signal x₂. Subset 2 (Ω₂) includes the points that relatively likely if the estimated modulation {circumflex over (M)}⁽²⁾ is not used by the interfering signal x₂.

The ratio of Equation 30 can be compared to a threshold to determine whether the observed distribution is incorrectly biased relative to the expected distribution of the assumed modulation {circumflex over (M)}⁽²⁾, as follows:ρ^(M2)>ρ_threshold

Equation 30A

The decision based on Equation 30A may be used to qualify or refine decisions regarding {circumflex over (M)}⁽²⁾ made using the Joint ML Sequence and Modulation Detection or sequential modulation and sequence detection schemes discussed above.

Subsampled Sequences

In the various techniques and methods described above, the interfering channel x₂ may be subsampled in order to reduce the number of calculations involved in any given technique. For example, r may be sampled once in every 100 symbols over a received sequence of 10000 symbols, in situations in which the modulation is anticipated to remain constant over at least 10000 symbols.

More particularly, the subsampling may be performed as follows:

$\begin{array}{cc}{\hat{M}}^{\left(2\right)}=\underset{M^{\left(2\right)},x_1,x_2}{\arg \max }{\left(\frac{1}{M^{\left(2\right)}}\right)}^L\hspace{1em}\prod _{j=1}^{L^{\prime }}\left(\sum _{m=1}^{M^{\left(1\right)}}\sum _{n=1}^{M^{\left(2\right)}}\exp (-\frac{{r\left(i\right)-h_1\left(i_j\right)x_1^{\left(m\right)}-h_2\left(i_j\right)x_2^{\left(n\right)}}^2}{{\sigma }^2})\right)& \mathrm{Equation}31\end{array}$ In addition to applying to log-MAP in Equation 31, subsampling can be applied to the technique of sum-max-log-MAP, mean-max-log-MAP, max-log-MAP-scale, log-MAP-bests-K similarly. Subsampling can be also be applied in combination with other techniques such as hard decision distribution metric discussed above, and adaptive ML/MMSE hereafter.Adaptive ML/MMSE

In certain embodiments, the ML detection schemes described above may be used when they produce reliable or accurate results, and MMSE may be used in other situations, where noise dominates or the modulation detection is deemed to be unreliable. Various measures or metrics may be used to determine whether the ML modulation detection is or is likely to be reliable and/or accurate. For example, signal-to-noise ratio (SNR) and/or SINR (signal-to-interference/noise ratio) of the interfering channel x₂ may be used as indicators of detection reliability.

As another example, the hard decision matrix metric described above, or any other observed quality metric, may be used in determining the actual accuracy or reliability of modulation detection. More specifically, the ratio of the minimum and maximum metrics may be evaluated and compared to a threshold to indicate the likely reliability of modulation detection, as follows:

$\begin{array}{cc}\theta =\frac{\underset{M^{\left(2\right)}\in \Psi }{\max }J\left(M^{\left(2\right)}\right)}{\underset{M^{\left(2\right)}\in \Psi }{\min }J\left(M^{\left(2\right)}\right)}\le {\theta }_0& \mathrm{Equation}32\end{array}$ Interference Cancellation without Modulation Knowledge

Typically, interference cancellation requires knowledge of the modulation of the interfering signal as well as coding information for the interfering signal. However, the following technique can be used in situations in which the modulation information of the interfering signal is not known at the receiver.

FIG. 3 shows an example signal flow, illustrating interference cancellation without knowledge of the modulation of the interfering signal. The signal r is received at block 302, which selects one of the available or possible modulations M⁽²⁾ for x₂. A MIMO receiver 304 is applied to the interfering signal x₂ and used as the basis for calculating the a-posteriori probability of the selected M⁽²⁾. The estimated probabilities for all available modulations are provided to a block 306 that calculations soft estimates of the interfering signal x₂. These are subtracted from the incoming signal r, and the resulting signal is provided to a MIMO receiver 308 for x₁, and a subsequent channel decoder 310 for x₁. If coding information is available for the interfering signal x₂, a channel decoder 312 may be used and its output provided to the soft estimation 306.

As an example of this process, assuming that coding information for the interfering signal x₂ is not available, a linear MMSE receiver may be applied to equalize the MIMO channel as follows:{tilde over (x)}₂=αh₂^H(HH^H+σ²I)⁻¹r=x₂+z

Equation 33

Based on Equation 31, the probability of each available modulation may be calculated as follows:

$\begin{array}{cc}P\left(M^{\left(2\right)}|r\right)=c·P\left(r|M^{\left(2\right)}\right)\approx c_1·\exp \left(-\frac{1}{L}\sum _{i=1}^L\frac{{{\tilde{x}}_2\left(i\right)-{\hat{x}}_2\left(i\right)}^2}{{\sigma }_z^2\left(i\right)}-\log \left(M^{\left(2\right)}\right)\right)=\frac{\exp \left(-\frac{1}{L}\sum _{i=1}^L\frac{{{\tilde{x}}_2\left(i\right)-{\hat{x}}_2\left(i\right)}^2}{{\sigma }_z^2\left(i\right)}-\log \left(M^{\left(2\right)}\right)\right)}{\sum _{M_2\in \Psi }\exp \left(-\frac{1}{L}\sum _{i=1}^L\frac{{{\tilde{x}}_2\left(i\right)-{\hat{x}}_2\left(i\right)}^2}{{\sigma }_z^2\left(i\right)}-\log \left(M^{\left(2\right)}\right)\right)}& \mathrm{Equation}34\end{array}$

It is to be noted that the Sum-Max-Log-MAP, as described above, is used in Equation 34 to approximate probabilities. Other described methods, such as Log-MAP, Max-Log-MAP-Scale, and so forth can also be applied here.

It is to also be noted that FIG. 3 is an example in which the probability of modulation is calculated after a linear receiver is applied to detect the interfering signal in Equations 33 and 34. The a-posteriori probability of the modulation can also be calculated directly from the received signals, as in the previous examples of Equation 19. Other described methods, such as Log-MAP, Max-Log-MAP-Scale, Sum/Mean-Max-Log-MAP can also be applied here.

The soft estimation for each hypothesized modulation may then be calculated as follows:

$\begin{array}{cc}E\left[x_2\left(i\right)|r,M^{\left(2\right)}\right]=\sum _{x_2\in {\Omega }^{\left(M^{\left(2\right)}\right)}}{x}_2·P\left(x_2|r\left(i\right),M^{\left(2\right)}\right)=\frac{\sum _{x_2\in {\Omega }^{\left(M^{\left(2\right)}\right)}}\left(x_2\exp \left(-\frac{{{\tilde{x}}_2\left(i\right)-x_2}^2}{{\sigma }_z^2\left(i\right)}\right)\right)}{\sum _{x_2\in {\Omega }^{\left(M^{\left(2\right)}\right)}}\exp \left(-\frac{{{\tilde{x}}_2\left(i\right)-x_2}^2}{{\sigma }_z^2\left(i\right)}\right)}& \mathrm{Equation}35\end{array}$

Again, the different kinds of Log-MAP techniques described above can be applied to Equation 35 to lower computational complexities of determining the probability of every possible modulated symbol in each modulation.

The soft estimation is then calculated over all possible modulations, as follows:

$\begin{array}{cc}{\hat{x}}_2\left(i\right)=E\left[x_2\left(i\right)|r\right]=\sum _{M^{\left(2\right)}\in \Psi }E\left[x_2|r,M^{\left(2\right)}\right]P\left(M^{\left(2\right)}|r\right)& \mathrm{Equation}36\end{array}$

Based on the soft estimation of Equation 36, the interference may be canceled from the interfering signal x₂, and the desired signal x₁ may be decoded as follows:{tilde over (r)}(i)=r(i)−h₂{circumflex over (x)}₂(i){circumflex over (x)}₁(i)

Equation 37

Both FIG. 2 and FIG. 3 involve determining a-posteriori probability of the modulation for the interfering signal. An example difference of the techniques of FIGS. 2 and 3 can be interpreted as a hard and soft detection of the modulation. The techniques in FIG. 2 and the associated variations, for example, make a hard decision on the modulation for the interfering signals based on a-posteriori probabilities and then rely on a modulation-dependent MIMO detection (e.g., ML detection) given the modulation of the interfering signals for interference suppression, or performance improvement. The techniques in FIG. 3 and the associated variations, for example, do not need to determine the modulation of the interfering signals. Instead, the expectation of the interfering signal, or a “soft” usage of the modulation of the interfering signals, is used for interference cancellation in FIG. 3.

A combination of hard and soft modulation detection is possible, where the modulation of the interfering signal is determined, and then a soft expectation of the interfering signals is cancelled from the received signal. This can be interpreted as another example of modulation dependent MIMO detection.

Another combination of soft and hard modulation detection is to find a hard estimate of modulated symbols in each modulation, and determine the expected interfering signals as the sum of each modulated symbols weighted by the probability of the modulation. This can be interpreted as the Sum-Max-Log-MAP technique for determining the probability of each modulated symbol given the modulation.

CONCLUSION

The discussion above assumes two rank-1 users for ease of description. However, the described techniques are not limited to the number of users or the ranks of the users. Rather, the described techniques can be extended to arbitrary numbers of users with arbitrary ranks, where the effective channels are available but the modulations of the channels are unknown at the receiver. For example, ML sequence and modulation detection with Sum-Max-Log-Map techniques may be performed with respect to multiple users in accordance with the following:

$\begin{array}{cc}\left\{{\hat{x}}_1\left(1\right)\dots {\hat{x}}_1\left(L\right),{\hat{M}}^{\left(2\right)}\dots {\hat{M}}^{\left(K\right)}\right\}=\hspace{1em}\underset{M^{\left(1\right)}\dots M^{\left(K\right)},x_1\dots x_K}{\arg \min }\left(\frac{1}{L}\sum _{i=1}^L\frac{\begin{array}{c}\min r\left(i\right)-H_1\left(i\right)x_1^{\left(m_1\right)}\left(i\right)-\\ {H_2\left(i\right)x_2^{\left(m_2\right)}\left(i\right)\dots -H_K\left(i\right)x_K^{\left(m_x\right)}\left(i\right)}^2\end{array}}{{\sigma }^2}+\sum _{k=2}^K\log \left(M^{\left(k\right)}\right)\right)& \mathrm{Equation}38\end{array}$

In the case of more than 1 interfering signals, the detection of the modulation for each signal can be done jointly by solving equation 38, for example if Sum-Max-Log-MAP technique is applied. Alternatively (or additionally), the modulation of each signal can be successively detected, treating all the signals with modulations undetected as noise. Another alternative is to detect the modulation of each signal in parallel, treating all the rest signals with modulations as noise. Furthermore, the joint and successive/parallel modulation detection for more than 1 interfering signals can be applied at the same time, for example, by grouping interfering signals, and applying the joint or successive/parallel principle to each group, and then applying the joint or successive/parallel principle within each group.

The discussion above also assume the existence of the interfering signals has been identified. Alternatively, the existence of the interfering signals can be included in the modulation detection by assigning the non-existence of the interfering signals as a special modulation and included in the set of modulations ψ. If the modulation of “non-existence” has been identified as the most likely one, the receiver may deem that the interfering signals are not present.

Note that the description above incorporates use of the phrases “in an embodiment,” or “in various embodiments,” or the like, which may each refer to one or more of the same or different embodiments. Furthermore, the terms “comprising,” “including,” “having,” and the like, as used with respect to embodiments of the present disclosure, are synonymous.

As used herein, the terms “logic,” “component,” and “module” may refer to, be part of, or include an Application Specific Integrated Circuit (ASIC), an electronic circuit, a processor (shared, dedicated, or group) and/or memory (shared, dedicated, or group) that execute one or more software or firmware programs, a combinational logic circuit, and/or other suitable components that provide the described functionality. The logic and functionality described herein may be implemented by any such components.

In accordance with various embodiments, an article of manufacture may be provided that includes a storage medium having instructions stored thereon that, if executed, result in the operations described above. In an embodiment, the storage medium comprises some type of non-transitory memory (not shown). In accordance with various embodiments, the article of manufacture may be a computer-readable medium such as, for example, software or firmware.

Various operations may have been described as multiple discrete actions or operations in turn, in a manner that is most helpful in understanding the claimed subject matter. However, the order of description should not be construed as to imply that these operations are necessarily order dependent. In particular, these operations may not be performed in the order of presentation. Operations described may be performed in a different order than the described embodiment. Various additional operations may be performed and/or described operations may be omitted in additional embodiments.

Although certain embodiments have been illustrated and described herein, a wide variety of alternate and/or equivalent embodiments or implementations calculated to achieve the same purposes may be substituted for the embodiments illustrated and described without departing from the scope of the present disclosure. This application is intended to cover any adaptations or variations of the embodiments discussed herein. Therefore, it is manifestly intended that embodiments in accordance with the present disclosure be limited only by the claims and the equivalents thereof.

Claims

1. A method, comprising: receiving, at a receiver, a desired signal and an interfering signal, wherein the interfering signal was transmitted with a modulation unknown to the receiver;identifying a likely modulation corresponding to the modulation with which the interfering signal was transmitted; anddecoding the desired signal using a modulation dependent multiple-input multiple output (MIMO) detection, wherein the modulation dependent MIMO detection is based at least in part on the identified likely modulation corresponding to the modulation with which the interfering signal was transmitted, wherein the modulation dependent MIMO detection includes maximum likelihood (ML) detection.

2. The method of claim 1, wherein identifying the likely modulation comprises arbitrarily selecting a modulation from a set of possible modulations.

3. The method of claim 1, wherein identifying the likely modulation comprises calculating a superposition of multiple possible modulations of the interfering signal.

4. The method of claim 1, wherein identifying the likely modulation comprises applying a linear receiver to the interfering signal to identify the likely modulation.

5. The method of claim 4, further comprising: detecting the interfering signal using the linear receiver multiple times, such that for each time of the multiple times, a corresponding possible modulation from a set of modulations is assumed while detecting the interfering signal using the linear receiver for the corresponding time; andmaximizing an a-posteriori probability of each of the possible modulations from the set of modulations.

6. The method of claim 5, wherein maximizing the a-posteriori probability of the possible modulation from the set of modulations and detecting the interfering signal using the linear receiver is based at least in part on assuming that the possible modulation is simplified by minimizing a minimum distance offset by a value that is based on one or both of the possible modulation and a residual noise.

7. The method of claim 5, wherein maximizing the a-posteriori probability of the possible modulation from the set of modulations and detecting the interfering signal using the linear receiver is based at least in part on assuming that the possible modulation is simplified by maximizing at least in part a sum of largest K conditional probabilities.

8. The method of claim 5, wherein an a-priori probability of the possible modulation is scaled at least in part in accordance with a distribution of the interfering signal detected by the linear receiver assuming the possible modulation.

9. The method of claim 1, wherein: identifying the likely modulation comprises decoding the interfering signal multiple times using different modulations from a set of possible modulations; anddetermining which of the modulations results in accurate decoding of the interfering signal.

10. The method of claim 1, wherein identifying the likely modulation and decoding the desired signal comprise maximizing an a-posteriori probability of the desired signal and a possible modulation from a set of modulations.

11. The method of claim 10, wherein maximizing the a-posteriori probability is simplified at least in part by minimizing a minimum distance offset by a value, and wherein the value is based on one or more of the possible modulation, a noise power, and a condition of a channel.

12. The method of claim 10, wherein maximizing the a-posteriori probability is simplified by maximizing at least in part a sum of the largest K conditional probabilities.

13. The method of claim 1, wherein identifying the likely modulation further comprises selecting at least in part a subset of the received signal.

14. The method of claim 1, wherein identifying the likely modulation further comprises: determining a reliability of the likely modulation identified; andreplacing the modulation dependent MIMO detection by a linear receiver.

15. The method of claim 1, wherein the interfering signal is a first interfering signal, and wherein the method further comprises: receiving, at the receiver, a plurality of interfering signals including the first interfering signal, wherein each of the plurality of interfering signals were transmitted with a respective modulation unknown to the receiver; andidentifying a plurality of likely modulations corresponding to a plurality of modulations with which the plurality of interfering signals were respectively transmitted,wherein the modulation dependent MIMO detection is based at least in part on the identified plurality of likely modulations corresponding to the plurality of modulations with which the plurality of interfering signals were respectively transmitted.

16. A system comprising: one or more antennas configured to receive a desired signal and an interfering signal, wherein the interfering signal was transmitted with a modulation unknown to the system; anda data processing unit coupled to the one or more antennas, wherein the data processing unit is configured to (i) identify a likely modulation corresponding to the modulation with which the interfering signal was transmitted, and (ii) decode the desired signal using a modulation dependent detection, wherein the modulation dependent detection is based at least in part on the identified likely modulation corresponding to the modulation with which the interfering signal was transmitted, wherein the modulation dependent detection includes maximum likelihood (ML) detection.

17. The system of claim 16, wherein the data processing unit is further configured to identify the likely modulation by arbitrarily selecting a modulation from a set of possible modulations.

18. The system of claim 16, wherein the data processing unit is further configured to identify the likely modulation by calculating a superposition of multiple possible modulations of the interfering signal.

19. The system of claim 16, wherein the data processing unit is further configured to identify the likely modulation by applying a linear receiver to the interfering signal.

20. The system of claim 16, wherein the data processing unit is further configured to identify the likely modulation by (i) detecting the interfering signal using the linear receiver multiple times, such that for each time of the multiple times, a corresponding possible modulation from a set of modulations is assumed while detecting the interfering signal using the linear receiver for the corresponding time, and (ii) maximizing an a-posteriori probability of each of the possible modulations from the set of modulations.

21. The system of claim 20, wherein the data processing unit is further configured to maximize the a-posteriori probability of the possible modulation and detect the interfering signals by: minimizing at least in part a minimum distance offset by a value, wherein the value is based on one or more of the possible modulation and a residual noise.

22. The system of claim 20, wherein the data processing unit is further configured to maximize the a-posteriori probability of the possible modulation and detect the interfering signals by: maximizing at least in part a sum of the largest K conditional probabilities.

23. The system of claim 20, wherein the data processing unit is further configured to scale an a-priori probability of the possible modulation based at least in part on a distribution of the interfering signal detected by the linear receiver assuming the possible modulation.

24. The system of claim 16, wherein the data processing unit is further configured to identify the likely modulation by (i) decoding the interfering signal multiple times using different modulations from a set of possible modulations and (ii) determining which of the modulations results in accurate decoding of the interfering signal.

25. The system of claim 16, wherein the data processing unit is further configured to identify the likely modulation and decode the desired signal by maximizing an a-posteriori probability of the desired signal and a possible modulation from a set of modulations.

26. The system of claim 25, wherein the data processing unit is further configured to simplify maximizing the a-posteriori probability at least in part by minimizing the minimum distance offset by a value, and wherein the value depends on one or more factors that are the possible modulation, the noise power, the channels.

27. The system of claim 25, wherein the data processing unit is further configured to simplify maximizing the a-posteriori probability by maximizing at least in part a sum of the largest K conditional probabilities.

28. The system of claim 16, wherein the data processing unit is further configured to select at least in part a subset of received signals to identify the likely modulation.

29. The system of claim 16, wherein the data processing unit is further configured to (i) determine a reliability of the likely modulation, and (ii) based on determine the reliability of the likely modulation, replace the modulation dependent MIMO detection with a detection using a linear receiver.