Patent Application
20250102620
The present invention generally relates to a method for estimating a center frequency of a wavelet, and an electronic device.
There is a growing interest in radar technology enabling short range and high precision measurements in industrial, medical, security as well as consumer applications. Non-limiting example applications include ranging and positioning, presence and gesture detection, and investigation of material properties such as thickness, size, dielectric properties, material composition etc.
In pulse-based radar systems the send time may be kept short, which in turn enables a lower power consumption. Ultra-wideband, UWB, pulses (e.g. with a bandwidth of a few GHz or more) may as an example allow pulse durations in the order of nanoseconds, hundreds of picoseconds or shorter. In a pulse-based radar system a radar pulse (“wavelet”) is transmitted to be reflected off a radar target, wherein the reflected radar pulse is received and analyzed, e.g. to determine a time-of-flight of the radar pulse, and/or other characteristics of the reflected pulse such as amplitude and phase. Pulse-based radar systems operating in the millimetre band, or mmWave band, (herein considered to correspond to the spectrum extending from 30 GHz to 300 GHz) are of particular interest as the short wavelength of mmWave radar wavelets facilitates high accuracy measurements in short range applications.
Precise on-chip measurements of the center frequency of generated radar wavelets is an important function of a radar device, both for facilitating precise measurements and for device calibration (e.g. to ensure complying with regulatory spectrum allocations). In order to sample the wavelet signal to be measured, the sampling circuit (e.g. the ADC) must have a sufficient sample rate, i.e. bandwidth, to be able to measure the signal S without undersampling or aliasing. This becomes increasingly challenging at the high frequencies of mmWave signals and large bandwidth of UWB signals. Although wideband ADCs do exist, they may not be feasible to implement in compact and power-efficient devices.
It is an object of the present invention to provide a method enabling center frequency estimation of a wavelet with reduced bandwidth requirements for the sampling circuitry, which may be implemented on-chip in compact and power-efficient devices. In particular, it is an object to enable such center frequency estimation also for UWB wavelets in the mmWave band.
Hence, according to a first aspect there is provided a method for estimating a center frequency of a wavelet, the method comprising:
According to a second aspect there is provided an electronic device comprising:
The present invention is based on the idea that rather than directly sampling the first wavelet whose frequency is to be measured, the center frequency may be estimated from a sequence of samples representing a portion of a convolution between the first wavelet and the second wavelet (“probe wavelet”). This in turn allows the sequence of samples to obtained in a time-diluted fashion, i.e. by repeatedly performing a “convolution measurement” while varying a timing offset (i.e. a relative delay) between the convolved signals, i.e. the first wavelet and the probe signal.
Underlying this approach is the mathematical relationship that a convolution of a first periodic signal A(t) of period T (i.e. A(t)=A(t+T)) with any other signal B(t) will yield a convolved signal C(t)=(A*B)(t) that is also periodic with the period T of the first signal. Accordingly, the frequency of the first signal x(t) may be obtained as the frequency f=1/T of the convolved signal z(t). While this relationship, in strict mathematical sense applies to infinitely periodic signals, it also holds true, at least approximately, to points ti of the convolved signal z(t=ti) for which the other signal B(t) (corresponding to the probe wavelet) overlaps only a portion of substantially constant frequency of the first signal A(t) (corresponding to the first wavelet). The portion of substantially constant frequency of the convolution may in the following for conciseness be referred to as the constant frequency portion or the middle portion of the convolution. According to the method and device of the first and second aspects, this is reflected in that the different timing offsets of the measurements are such that each sample represents a respective point (i.e. time instant) of the portion of substantially constant center frequency of the convolution. Hence, the initial and end portions of the first wavelet, which for typical wavelet generators tend to be characterized by unstable center frequency, are excluded from the convolution measurements used to obtain the sequence of samples on which the frequency estimation is based.
Moreover, in view of the above, the frequency estimation need not rely on precise control or knowledge of the characteristics of the probe wavelet, other than in ensuring that for none of the samples does the second/probe wavelet temporally overlap the inner or end portions of the first wavelet.
Thus, having obtained the sequence of samples representing a portion of substantially constant center frequency of the convolution, the center frequency of the first wavelet may be estimated by performing phase unwrapping and determining the time rate of change of the resulting sequence of un-wrapped phase values.
By the term “wavelet” is here meant a pulsed electro-magnetic oscillating signal having a carrier frequency (the center frequency of the wavelet) and a limited temporal extension, each wavelet including a plurality of oscillations or periods within the limited temporal extension. That is each “pulse” comprises a plurality of oscillations or periods in accordance with the carrier/center frequency of the wavelet.
By the term “sample of a convolution” (e.g. with reference to the first wavelet and the second/probe signal) is here meant a sample (digital or analog) of a given point of a waveform of the convolution. Due to the time-diluted sampling approach, the waveform of the full convolution need not as such be produced and then sampled, but rather a single sample representing a given point of the waveform may be obtained in each convolution measurement.
By the term “unwrap” with respect to phase values is here meant transforming the relative instantaneous phase values (wrapped from −π to π) of the waveform represented by the sequence of samples to absolute instantaneous phase values (by adding multiples of ±2π to the relative instantaneous phase values).
In some embodiments of the first method aspect, each measurement of the sequence of measurements comprises:
Correspondingly, in some embodiments of the second device aspect, the electronic device further comprises a mixer and integrator circuit, and wherein each measurement comprises:
By the term “mixing” is here meant RF mixing/frequency mixing (e.g. by a mixer portion of the mixer and integrator circuit) of two signals (e.g. the first wavelet and the second signal) to obtain a mixing product.
By the term “integrating” is here meant an integration (e.g. by an integrator portion of the mixer and integrator circuit) of a signal (e.g. the mixing product of the first wavelet and the second signal) to obtain a time integral of the signal (e.g. an integrated mixing product). The integration hence converts the mixing product of each measurement into a respective integrated mixing product. The integrated mixing product may in particular be a DC signal. The respective mixing product should be integrated over its full duration. That is, the time period during which the mixing product is accumulated (e.g. by the integrator) should cover the full duration of the mixing product.
By the term “mixer and integrator circuit” is here meant any circuit portion allowing frequency mixing and integration of the mixing product. A mixer and integrator circuit may comprise a discrete circuit portion providing the mixing and a discrete integrator portion coupled to an output of the mixing circuit portion and providing the integration. However, a mixer and integrator circuit may also be realized by a common circuit portion or element providing, simultaneously, the mixing and integration.
Each sample may thus be obtained by simply mixing and integrating the first wavelet and the second wavelet, with a given timing offset between the first and second wavelets. By varying the timing offsets between the measurements, samples representing different points or time instants of the waveform of the convolution may thus be obtained. E.g., by varying the timing offset between the first through fourth radar measurements, a relative time of arrival of the first wavelet at the mixer and integrator circuit, with respect to a time of arrival of the second wavelet at the mixer and integrator circuit may be varied between the measurements.
The measurements of the sequence of measurements are here to be understood as separate and sequential measurements, each comprising: generating a respective instance of the first wavelet and a respective instance of the second wavelet; mixing and integrating the respective instances of the first and second wavelets to obtain a respective integrated mixing product, and sampling the respective integrated mixing product to obtain the respective sample.
In some embodiments, the timing offset between the first and second wavelets is varied between the sequence of measurements such that the second wavelet presents a different overlap with the middle portion of the first wavelet in each measurement. Each sample may thus represent a different point of the convolution.
In some embodiments, the timing offset is varied over a range such that the sequence of samples spans at least one period of the convolution.
Basing the frequency estimation on samples spanning one or more periods of the convolution may contribute to the reliability and accuracy of the measurement.
In some embodiments, estimating the sequence of instantaneous phase values comprises estimating a sequence of in-phase and quadrature-phase components from the sequence of samples.
In some embodiments, the middle portion of the first wavelet further has a substantially uniform amplitude. This may improve the reliability and simplicity of the frequency estimation by reducing the number of variable parameters in the measurements. Amplitude variations in the middle portion of the first wavelet may otherwise tend to increase the uncertainty of the frequency estimation as the modulation may result in a broadening of the carrier frequency.
In some embodiments, a center frequency of the first wavelet is unstable in the initial portion and/or end portion. As the sequence of samples on which the center frequency estimation is based are obtained from the convolution measurements from the middle portion of the first wavelet, frequency instability in the initial and/or end portion will not adversely affect the frequency estimation. A related advantage is that the requirements on the frequency stability of the first wavelet generator during start-up and shut-down may be reduced.
In some embodiments, the first wavelet generator has a configurable duration. Hence, the duration of the first wavelet may be increased for the purpose of facilitating the frequency estimation, and after the frequency has been estimated, be reduced to a duration suitable for the particular application (e.g. radar measurement).
In some embodiments, a frequency spectrum of the second wavelet comprises a band overlapping the center frequency of the first wavelet. This will reliably produce a peak in a frequency distribution of a convolution between such first and second wavelets, at a frequency matching the center frequency of the first wavelet.
In some embodiments, the second wavelet presents a substantially triangular or trapezoidal envelope, or the second wavelet comprises a frequency-chirp. These example implementations allows providing a second wavelet with a relatively broad frequency band of non-zero energy, to be mixed with the first wavelet, thus facilitating providing an overlap with the center frequency of the first wavelet.
In some embodiments, the method further comprises tuning a resonance frequency of the first wavelet generator responsive to the estimated center frequency deviating from a reference frequency or reference frequency band.
The frequency estimate may thus be used for adjusting the resonance frequency of the first wavelet generator so as to produce the first wavelets with a center frequency meeting the reference frequency or falling within the reference frequency band (e.g. to provide a desired measurement accuracy and/or for spectrum compliance).
In some embodiments, the first and second wavelet generators are comprised in a radar device.
In some embodiments, the first wavelet generator is configured to generate a wavelet for transmission from the radar device.
This and other aspects of the present invention will now be described in more detail, with reference to the appended drawings showing embodiments of the present invention.
The electronic device 1 comprises a first wavelet generator 6. The first wavelet generator 6 is configured to generate a first wavelet W1 which is the wavelet whose center frequency is to be estimated. The electronic device 1 further comprises a second wavelet generator 8. The second wavelet generator 8 is configured to generate a second wavelet W2. As set out herein, the second wavelet W2 is used as a probe wavelet to facilitate the estimation of the center frequency first wavelet W1.
To facilitate the frequency estimation, the first and the second wavelet generators 6, 8 may each be respective coherent wavelet generators. That is, the first wavelet generator 6 may be configured to coherently generate instances of the first wavelet W1 such that the first wavelet generator 6 may generate substantially uniform instances of the first wavelet W1. By substantially uniform instances of the first wavelet W1 is here meant that the instances have substantially matching waveforms, e.g. in terms of initial phase, frequency, bandwidth and amplitude. Instances of the first wavelet W1 generated by first wavelet generator 6 at different points in time may thus present phase-coherence with respect to each other. This discussion applies correspondingly to the second wavelet generator 8 and the instances of the second wavelets W2. For convenience, the label W1 may in the following be used both with reference to the (waveform of the) first wavelet as well as instances thereof. Correspondingly, the label W2 may in the following be used both with reference to the (waveform of the) second wavelet as well as instances thereof.
The first and second wavelet generators 6, 8 may each be mmWave wavelet generators, i.e. be capable of generating wavelets with center frequencies in the mmWave band. Non-limiting examples of frequency bands within which the center frequencies of the wavelets W1, W2 may be generated include 57-71 GHz, 57-64 GHz and 77-81 GHz. The first and second wavelet generators 6, 8 may further be capable to generate the wavelets W1, W2 as UWB wavelets, e.g. with bandwidths of about 1 GHz to about 14 GHz (e.g. measured at FWHM). It is here to be noted that a capability of a wavelet generator to generate UWB wavelets does not preclude the possibility of generating smaller bandwidth (i.e. non-UWB) wavelets. As further will become apparent from the following discussion, to facilitate the frequency estimation of the first wavelet W1, the duration of the second wavelet W2 is smaller than the duration of the first wavelet W1, more specifically shorter than the duration of a middle/constant frequency portion of the first wavelet W1. Hence, the bandwidths of the first and second wavelets W1, W2 may be different. It is further to be noted that, while the ability to generate UWB wavelets (e.g. mmWave wavelets) may be useful in the particular application for which the electronic device 1 is intended to be used (e.g. for high accuracy measurements in short range radar applications, discussed below), the first and second wavelets W1, W2 need not for the purpose of the frequency estimation be UWB wavelets. For instance, as will be further elaborated upon below, a duration of wavelets generated by the first wavelet generator 6 may be configurable, wherein the first wavelet generator 6 may be configured to generate the first wavelet W1 of duration τ1 (e.g. a UWB or non-UWB wavelet) when performing frequency estimation and generate a third wavelet W3 of a same center frequency as the wavelet W1 but a shorter duration 13<11 (e.g. a UWB wavelet) when performing radar measurements.
The first and second wavelet generators 6, 8 may be implemented in various manners. According to one example, the wavelet generation may be realized using a transceiver module as described in connection with FIG. 4 of the published PCT-application PCT/EP2011/058847 on page 11 lines 22-30. This implementation offers a low-complexity method for phase-coherent generation of UWB mmWave wavelets with a well-defined and fixed initial phase, a configurable duration/bandwidth, and substantially constant carrier frequency (at least after a transient initial period). The wavelet generation may however also be realized using other oscillator configurations, for example, by a pulse generator relying on negative differential conductance (e.g. provided by a cross-coupled differential pair) or positive feedback, where it is possible to start and quench the oscillations rapidly. Further example implementations include passing a continuous oscillation through a switchable amplifier, filtering of a baseband pulse, up-conversion of a baseband pulse, or using digital gates to create the desired waveform, as per se is known to the person skilled in the art.
The first and second wavelet generators 6, 8 may as indicated in
The electronic device 1 may as shown further comprise a timing circuit 4. The timing circuit 4 is configured to generate and output the first and second trigger signals T1 and T2 with a controllable relative timing or timing offset Δ. The relative timing of the generation of the first and second wavelets W1, W2 may thus be controlled by controlling a relative timing of the trigger signals T1 and T2.
The timing circuit 4 may operate on basis of a reference clock Ref, e.g. provided by a voltage-controlled oscillator (VCO) or some other conventional clock source of the electronic device 1, or provided to the electronic device 1 from an external clock source. The timing circuit 4 may for example comprise counter circuitry configured to output the first trigger T1 and the second trigger T2 after a respective number of clock edges of the reference clock Ref.
Various implementations of trigger generating circuitry are possible, such as counter circuitry for counting edges of the clock signal Ref and generating the trigger signals T1 or T2 responsive to a trigger condition. The first and second trigger signals T1, T2 may for instance be generated by a respective Finite-State Machine (FSM), each comprising a counter circuit configured to count clock edges (e.g. leading and/or trailing) of the clock signal Ref and in response to a configurable number of clock edges (e.g. once every N edges received) provide the respective trigger signal T1 or T2. The FSM may comprise a register for storing state variables (e.g. the number N of clock edges defining the trigger condition, a pulse duration, etc.). The FSM may comprise a number of control inputs for receiving control signals (e.g. digital control words) for controlling the FSM and updating the register. An FSM may as per se is known in the art be implemented using a programmable logic device.
Higher precision delay control may be obtained by distributing the clock signal Ref along a first path associated with the first trigger signal generation T1 and a second path associated with the second trigger signal generation T2. A programmable delay (e.g. a delay line such as a tapped delay line, or some other conventional digital or continuous delay circuit) may be provided along one of the first or second paths to provide fine delay control between the respective clock signals propagating along the first and second paths. By way of example, state of the art delay circuits, such as tapped delay lines, may enable variable delay control with a resolution/step-size of about 0.5 ps or less. It is also possible to use separate and synchronized clocks sources for controlling the generation of the first and second triggers T1 and T2.
Still with reference to
The integrator 14 is meanwhile configured to integrate the mixing product M to output an integrated mixing product I. An input of the integrator 14 is coupled to the output of the mixer 12. The mixing product M (which for overlapping W1 and W2 instances will be an AC signal) may hence be converted to a representative DC signal or DC level. The integrator 14 may be implemented as a capacitive circuit, or an integrating OP-amp. In any case, an integration time of the integrator 14 may be configured to be equal to or greater than a maximum expected duration of the mixed wavelet of the mixing product M.
While in
The electronic device 1 further comprises a sampling circuit, which in the illustrated example is implemented as an analog-to-digital converter (ADC) 16. An input of the ADC 16 is coupled to the output of the mixer and integrator circuit 10 (e.g. the output of the integrator 14). The ADC 16 is configured to convert each DC-level integrated mixing product I to a digital sample x.
The electronic device 1 further comprises a signal processing circuit which in the illustrated example is implemented by a digital signal processor (DSP) 18. An input of the DSP 18 is coupled to an output of the ADC 16. As will be further described in connection with
Although not shown in
The method 20 comprises performing a sequence of measurements j=1 . . . M, wherein each measurement j comprises:
Thus, in each measurement j, a sample xj representing a respective point of a convolution of the first and second wavelet W1, W2 is obtained. The samples xj may be output to the DSP 18 such that the DSP 18 iteratively obtains a sequence of samples xj=1 . . . M to be further processed as set out in the following.
Assuming the propagation of the instances of the first and second wavelets W1, W2 between the respective wavelet generators 6, 8 and the mixer and integrator circuit 10 involves no losses or signal spread, and assuming the mixing and integration by the mixer and integrator circuit 10 is perfect, each sample x; is given by:
While the first wavelet generator 6 is able to coherently generate instances of the first wavelet W1 (i.e. with a well-defined and fixed initial phase), the center frequency may during the initial and/or end portions δi, δe be unstable, e.g. due to characteristics of the first wavelet generator 6. Therefore, as previously discussed, the samples of the convolution which may form the basis for a reliable and accurate center frequency estimate are those which are obtained by mixing the second wavelet W2 with the middle/constant frequency portion δm of the first wavelet W1. Accordingly, the duration δ2 of the second wavelet W2 is smaller than the duration δm of the middle/constant frequency portion δm of the first wavelet W1.
Furthermore, the timing offset Δj between the first and second wavelets W1, W2 is varied between the measurements j such that each sample xj of the sequence of samples xj=1 . . . M represents a respective point Δj within a portion of substantially constant center frequency (“constant frequency portion”) of the convolution (W1*W2). The samples xj are thus obtained when the timing offset Δj between the first and second wavelets W1, W2 is such that the second wavelet W2 overlaps the constant frequency portion δm of the first wavelet W1, but neither the initial portion δi nor the end portion de. The timing offset Δj between the first and second wavelets W1, W2 may be varied between the sequence of measurements j=1 . . . M such that the second wavelet W2 presents a different overlap with (i.e. overlaps a different sub-portion of) the constant frequency portion δm of the first wavelet W1 in each measurement. With reference to
As discussed previously, the periodicity (T3) and frequency (f3=1/T3) of the constant frequency portion 83 of the convolution (W1*W2) will be substantially the same as the periodicity (T1) and frequency (f1=1/T1=f3) of the constant frequency portion δm of the first wavelet W1. The sequence of samples xj=1 . . . M thus enables the center frequency of the first wavelet W1 to be estimated, as further will be discussed below.
The number of samples obtained from the constant frequency portion δ3 (i.e. corresponding to the number of times the convolution measurements are iterated) typically depends on factors such as desired measurement accuracy, amount of processing resources available for storing and processing the samples, and the time available for performing the measurements. To obtain a reliable frequency estimate, it may however typically be beneficial that the sequence of samples xj=1 . . . M spans at least one period of the constant frequency portion δ3 of the convolution. A greater span may both facilitate determining the time rate of change of the un-wrapped phase values (see below) and allow for averaging of the frequency estimate across a plurality of periods.
Correspondingly, the sample rate may depend on similar factors (e.g. measurement accuracy, available processing resources) but also on the demodulation approach used to estimate the phase. It is noted that due to the time-diluted sampling approach, the sample rate does here not correspond to the temporal rate (in real-time) at which the samples xj are obtained, but is rather given by the spacing of successive samples xj, xj+1 which is determined by the size of the time steps of Δj of the sequence of measurements j=1 . . . M. It is further contemplated that a non-uniform sampling rate may be used, wherein the sequence of samples xj=1 . . . M may comprise a number of sub-sequences of samples, wherein a spacing (i.e. the time steps Δj) between the samples of the respective sub-sequences is smaller than a spacing between the sub-sequences. As an example, the sequence of samples may span at least one period (advantageously at least two) of the constant frequency portion δ3 of the convolution and be acquired with a sample rate sufficient to avoid aliasing.
While, as discussed above, the samples relevant for the frequency estimation are those which represent/are obtained from the constant frequency portion 83 of the convolution, the method does not preclude obtaining samples also from either of the transient portions of the convolution (dashed line waveform in
After the samples xj=1 . . . M have been obtained, the method 200 proceeds by the DSP 18 estimating a sequence of instantaneous phase values ¢j=1 . . . K from the sequence of samples xj=1 . . . M (step S22). Estimating the instantaneous phase values ϕj=1 . . . K may comprise estimating a sequence of K in-phase (I) and quadrature-phase (Q) components (i.e. IQ-pairs) from the sequence of samples ϕj=1 . . . K. An instantaneous phase value may be estimated for each IQ-pairs (e.g. as the argument of the IQ-pair, arg (I,Q)). Any conventional IQ demodulation technique may be used. As would be understood by the skilled person, the number of K of instantaneous phase values obtained will depend on the type of IQ demodulation technique and the number of samples xj=1 . . . M.
The method thereafter proceeds by the DSP 18 subjecting the sequence of instantaneous phase values ϕj=1 . . . K to phase-unwrapping to obtain a sequence of un-wrapped phase values ψj=1 . . . L (step S23). The number of un-wrapped values ψj=1 . . . L may typically be equal to the number of instantaneous phase values ϕj, i.e. L=K. Phase-unwrapping may for instance, as per se is known in the art, be implemented by adding multiples of ±2π when consecutive instantaneous phase values of the sequence ϕj=1 . . . K is equal or greater than π to reduce the difference to less than π.
Subsequent to estimating the estimating the sequence of un-wrapped phase values ψi=1 . . . L the DSP 18 estimates the center frequency f1 of the first wavelet W1 by determining the time rate of change of the sequence of un-wrapped phase values ψj=1 . . . L. As an example, the center frequency f1 of the first wavelet W1 in the example of
The DSP 18 may store the frequency estimates in a memory of, or connected to, the DSP 18 for retrieval and/or further processing, and/or output to peripheral circuitry via a communication bus of the electronic device 1. As an example, the DSP 18 may compare the frequency f1 of the first wavelet W1 to a reference frequency or reference frequency band to determine whether the frequency f1 of the first wavelet W1 is within a tolerance.
If the electronic device 1 comprises a first wavelet generator 6 with a tunable resonance frequency (e.g. the carrier of the first wavelet W1 may be generated by a voltage-controlled oscillator), the resonance frequency may be tuned responsive to the estimated center frequency deviating from the reference frequency or reference frequency band such that the center frequency f1 is tuned within a tolerance of the reference frequency or band. This may be performed as an iterative process (e.g. iteratively: estimating f1, comparing f1 to the reference frequency or band, and tuning the resonance frequency) wherein the center frequency f1 may be tuned to iteratively approach the reference frequency or band (e.g. approach the center frequency of the reference band).
As mentioned above, the intermediate portion δm of the first wavelet W1 presents a substantially constant frequency. The term “substantially constant” may here be understood in light of the fact that even an oscillatory signal with a limited temporal extent will have a certain non-zero width in the frequency domain. Hence, even between the transient portions &t and δe of the first wavelet W1, in the intermediate portion δm the center frequency f1 may only be determined within a certain precision which in the end is limited among others by the duration δm. Accordingly, the peak in the spectrum of the first wavelet W1 which corresponds to the center frequency may have a non-zero bandwidth. In some implementations, the center frequency may vary by at most 5%, or at most 2% or at most 1% during the intermediate portion δm. Assuming the first wavelet W1 is a mmWave wavelet, the center frequency f1 may for example lie within a band of width 1 GHz or less, such as 0.5 GHz or 0.1 GHz. This discussion applies correspondingly to the substantially constant frequency portion of the convolution (W1*W2).
As further mentioned above, the intermediate portion δm of the first wavelet W1 may also present a substantially constant amplitude. This may improve the precision of the frequency estimation since an amplitude modulation within the intermediate portion δm of the first wavelet W1 will manifest itself as a (further) broadening of the peak corresponding to the center frequency f1. In general, an amplitude variation of at most 5%, or at most 2% or at most 1% may be considered as substantially constant within the measurement tolerances of the present frequency estimation method. That said, even greater amplitude variations may be tolerable in some instances, e.g. if there are greater error sources within the measurement which anyway mask the broadening due to the amplitude modulation.
To ensure the non-overlapping relationship of the second wavelet W2 with the leading and trailing portions δi and δe of the first wavelet W1 for all samples of the sequence xj=1 . . . K, the duration δ2 of the second wavelet W2 is smaller than the duration of the middle portion δm of the first wavelet W1 by an amount such that that the second wavelet W2 does not overlap the leading and trailing portions δi and δe for any of the timing offsets Δj=1 . . . K. However, there is in no requirement on a substantially constant center frequency or amplitude for the second wavelet W2. Accordingly, various designs of the second wavelet W2 are compatible with the frequency estimation method. However, it has been realized that it may be advantageous that the frequency spectrum of the second wavelet W2 comprises a lobe or peak overlapping (i.e. comprising) the center frequency f1 of the first wavelet W1 which is to be estimated. The lobe or peak may advantageously be a main lobe or main peak of the frequency spectrum of the second wavelet W2 as this allows the portion of the spectrum of the second wavelet W2 comprising the greatest energy to be mixed into the samples xj. However, also side lobes or secondary peaks may be used provided they include a substantial energy.
Common for the examples of
Referring again to
The frequency estimation method set out above may be used in various applications, including various types of electronic devices. Referring again to
The first wavelet generator 6 may for instance form part of a transmitter of the radar device 1. In a radar measurement, the radar device 1 may transmit wavelets generated by the first wavelet generator 6. The wavelets may be transmitted towards a radar target via an antenna 2 of the radar device 1 (e.g. a directional antenna such as a dipole antenna, a Yagi antenna or a phased array antenna). Radar returns reflected by a radar target may be received and processed by a receiver of the radar device 1 using techniques which per se are known in the art. For instance, radar returns comprising reflected wavelets may be received (e.g. via the antenna 2 or via a separate antenna of the receive part) and sampled. The radar samples may be processed to measure parameters such as time-of-flight, amplitude and/or phase. The radar returns may be sampled using any suitable existing techniques, such as time-diluted sampling of a sequence of radar returns from a sequence of coherently generated wavelets, or direct/real-time sampling of each radar return (provided the sampling circuit of the receiver has a sufficient bandwidth).
In a gesture detection application the radar target may be a hand or a finger of a user locate in relative proximity to an electronic device (e.g. a tablet computer, a laptop computer, a media player, a smart watch, headphones or some other electronic device implementing a radar-based contactless user interaction). In an autonomous robot application (e.g. a cleaning robot, a lawn mower robot, a drone), the radar target may be a structure such as a wall or some obstacle whose presence and/or position may be relevant for a movement decision by the autonomous robot. In a surface scanning application the radar target may be a surface of interest (e.g. a surface traversed by a robot). These examples of applications and radar targets are however merely non-limiting examples and many other types of applications and radar targets are possible.
The first wavelet generator 6 may also or alternatively form part of a receiver of the radar device 1, wherein the first wavelet generator 6 may generate reference or template wavelets for use in time-diluted sampling of radar returns.
The second wavelet generator 8 may on the other hand have a function only, or at least, in connection with the wavelet frequency estimation of the first wavelets W1. That is, the second wavelet generator 8 need not be used during actual radar measurements but only when the frequency of the first wavelet W1 is to be estimated.
The person skilled in the art realizes that the present invention by no means is limited to the embodiments and examples described above. On the contrary, many modifications and variations are possible within the scope of the appended claims. For example, while reference in the above was made specifically to radar applications, the methods of the present disclosure may be used also in transmitters or other signal processing systems, such as data modulation in communication devices.
| Number | Date | Country | Kind |
|---|---|---|---|
| 23199885.7 | Sep 2023 | EP | regional |
