Patent Application
20110231155
Embodiments of the present technique relate generally to sampling techniques, and more particularly to a system and method for compressively sampling a signal of interest.
Success of digital data acquisition processes has placed enormous pressure on signal processing hardware and software to support higher resolutions, denser sampling, a large number of sensors and an even larger number of modalities. Conventionally, digital data acquisition processes employ the Nyquist-Shannon sampling theorem that provides uniform sampling of data at the Nyquist rate, that is, at twice the bandwidth. However, most signals are sparse and contain several coefficients close to or equal to zero when represented in a linear transform domain, such as, frequency, wavelet or time. Therefore, sampling these sparse signals at the Nyquist rate, which is a worst-case threshold for any band-limited data, results in oversampling of the signal. This oversampling may further result in unnecessary computation, storage and battery requirements, thereby severely limiting the capabilities and performance of digital devices such as cameras, microarrays and wireless sensor networks.
Compressive sensing (CS) is an emerging field that provides a framework for efficient sampling of sparse signals using sub-Nyquist sampling rates. By employing CS, a sparse signal can be perfectly reconstructed, or robustly approximated, from a small set of random projections even in the presence of noise with sub-Nyquist sampling rates. Particularly, CS exploits a priori signal sparsity information for estimating signals in the presence of noise and solving signal restoration and imaging problems. Moreover, each compressively sampled measurement may include substantially the same amount of information, thereby simplifying the encoding and quantization processes.
Compressive sensing, therefore, has been applied in a variety of technology areas such as inventory management, homeland security, healthcare, Magnetic Resonance Imaging (MRI), and geo-sensing applications. Most CS systems, however, are customized for specific application requirements with each CS component being custom built to perform a specific set of functions. Such customization burdens the available space, power and computational resources of devices using multiple sensors that sample multiple signals for implementing feature-rich applications. Moreover, use of these customized components limit scalability and adaptability of the CS systems. Additionally, such configurations fail to allow updates to existing functions or dynamic mitigation of detected software and hardware errors.
It may therefore be desirable to develop a generic sampling technique for compressively sampling a plurality of signals even in the absence of prior knowledge or assumptions about the signals and corresponding applications. Particularly, there is a need for an adaptive system and method for dynamically configuring CS protocols based on a specified set of parameters for implementing desired functions and achieving a desired sampling performance.
In accordance with aspects of the present technique, a method for configuring a sensor chassis is presented. The method includes remotely receiving a set of parameters for compressively sampling an input signal. Further, a CS protocol for compressively sampling the input signal may be dynamically determined based on the remotely received set of parameters for achieving a desired sampling performance. Subsequently, the input signal is compressively sampled according to the determined CS protocol.
In accordance with a further aspect of the present technique, a sensor chassis is disclosed. The sensor chassis includes a receiver that receives an input signal and a processing subsystem that remotely receives a set of parameters for compressively sampling the input signal. Further, the processing subsystem may dynamically determine a CS protocol for compressively sampling the input signal based on the remotely received set of parameters for achieving a desired sampling performance. To that end, the sensor chassis may include one or more programmable filters, where each programmable filter has at least one setting whose value may be adjusted according to the determined CS protocol. Subsequently, the sensor chassis compressively samples the input signal according to the determined CS protocol.
These and other features, aspects, and advantages of the present technique will become better understood when the following detailed description is read with reference to the accompanying drawings in which like characters represent like parts throughout the drawings, wherein:
The following description presents a technique for dynamically configuring a sensor chassis for compressively sampling an input signal. Particularly, embodiments illustrated hereinafter describe a sensor chassis and a method for dynamically configuring the sensor chassis to compressively sample the input signal based on one or more received parameters. Although the following description includes only a few embodiments, the present technique may be implemented in many different operating environments and systems for compressively sampling a plurality of signals of interest. By way of example, the present technique may be used in environment monitoring, inventory management, homeland security, healthcare, Magnetic Resonance Imaging (MRI), and wireless sensing applications. An exemplary environment that is suitable for practising various implementations of the present technique will be discussed in the following sections with reference to
Additionally, the set of parameters may also include a parameter corresponding to the sensor chassis 102 (sensor chassis parameter) and a criterion specifying the desired sampling performance. By way of example, the sensor chassis parameter may include a type of analog-to-digital converter (ADC) to be used, a sampling rate, a desired number of bits per sample, or combinations thereof. Moreover, the desired sampling performance may correspond to a maximum acceptable difference between the first input signal 106 and a signal reconstructed according to the determined CS protocol. As used herein, the term “maximum acceptable difference” is defined as a reconstruction not differing from the first input signal 106 by more than a determined amount in a voltage or a power domain.
In accordance with aspects of the present technique, the set of parameters may be transmitted to a computing device 112 communicatively coupled to the sensors 104 and 108 over a communication network 114. The communication network 114 may include either or both of wired networks such as LAN and cable, and wireless networks such as WLAN, cellular networks, and/or satellite networks. Particularly, the set of parameters may be remotely received by the computing device 112 over the communication network 114. These parameters may generally be referred to as a remotely received set of parameters. As used herein, the term “remotely received set of parameters” refers to the set of parameters that may be indirectly received by the computing device 112 through a receiver 116 operatively coupled to at least one of the sensors 104 and 108, a user interface 118, a digital communication link 120 or a data repository 122 coupled to the computing device 112 over the communication network 114. In certain embodiments, however, the term “remotely received set of parameters” refers to the set of parameters that may be indirectly received by the sensor chassis 102 through the receiver 116 operatively coupled to at least one of the sensors 104 and 108, the user interface 118, the digital communication link 120 or the data repository 122 over the communication network 114.
Further, in accordance with aspects of the present technique, the computing device 112 may evaluate the remotely received set of parameters to determine one or more characteristics corresponding to the first input signal 106 and desired application and/or user requirements. For example, in case of a fire that originates in the particular region, one or more obstructions may block certain paths to an exit. The computing device 112 may evaluate the set of parameters received during a particular time interval from the sensors 104 and 108 positioned in and around the particular region. Particularly, the computing device 112 may evaluate a change in temperature that may be detected by the second sensor 108 to efficiently locate the fire. Additionally, the computing device 112 may also evaluate a change in determined positions of one or more objects in the particular region detected by the first sensor 104 to ascertain if objects have moved to create obstructions to the exit. The evaluation, thus, may allow security personnel to locate and evacuate people quickly and efficiently. To that end, the computing device 112 may include a processor 124 and a memory 126 for evaluating the received set of parameters. By way of example, the processor 124 may include one or more microprocessors, microcomputers, microcontrollers, dual core processors, and so forth. The processor 124 may dynamically determine a CS protocol for compressively sampling the first input signal 106 based on the remotely received set of parameters. Particularly, the processor 124 may evaluate the set of parameters to determine the CS protocol that may be used by the sensor chassis 102 to compressively sample the first input signal 106 to achieve the desired sampling performance.
In accordance with a further aspect of the present technique, the processor 124 may store one or more instructions corresponding to the determined CS protocol on a storage device coupled to the computing device 112. In a presently contemplated configuration, the processor 124 may store the one or more instructions corresponding to the determined CS protocol on a sampling control unit 128. In such a configuration the sampling control unit 128 may be an independent unit physically removed from the computing device 112 and/or the sensor chassis 102. In one embodiment, the independent sampling control unit 128 may initially be communicatively coupled to the computing device 112 for facilitating the processor 124 to program and store the one or more instructions on the sampling control unit 128. Subsequently, the sampling control unit 128, thus programmed by the processor 124, may be communicatively coupled to the sensor chassis 102 for compressively sampling the first input signal 106 based on the determined CS protocol. In accordance with aspects of the present technique, the sampling control unit 128 may include at least one of a memory device, a programmable device, and/or instructions received through a control device operatively coupled to the sensor chassis 102. Particularly, in one implementation, the sampling control unit 128 may include a field programmable gate array (FPGA). The FPGA implementation may allow dynamic configuration of multiple CS protocols, thus providing immense scalability and adaptability to the sensor chassis 102. Alternatively, the sampling control unit 128 may be implemented as an optical disk, a tape, a compact disk, and so on. The exemplary implementation, thus, may enable fabrication of a generic sensor chassis that may be configured ‘on the fly’ to dynamically select an appropriate CS protocol for sampling any received input signal. Such a generic sensor chassis may reduce the time and complexity involved in manufacturing and operating the sensor chassis. Additionally, the generic sensor chassis may also facilitate sampling of a plurality of input signals based on the structure of the input signals and ambient conditions.
Turning to
In accordance with aspects of the present technique, the processing subsystem 214 may use one or more parameters corresponding to the sampled input signal 106 to monitor the sampling performance of the sensor chassis 202. In case the desired sampling performance is not achieved, the sensor chassis 202 may provide an alert through an output device 218 coupled to the sensor chassis 202. Subsequently, in certain embodiments, the processing subsystem 214 may further customize the determined CS protocol to achieve the desired sampling performance upon receiving the alert through the output device 218. By way of example, the output device 218 may include visual indicators such as a display and blinking lights, audio indicators such as speakers, and so on. Additionally, the sensor chassis 202 may include a power source 216 for operating the sensor chassis 202. The power source 216 may include a battery, line power, solar or wind powered cells, and so on to suit desired application and deployment needs. By way of example, in an air sampling system, the sensor chassis 202 may use a solar powered cell as the power source 216, whereas in a deep-sea sampling system a lead-acid battery may be used as the power source 216.
Thus, the sensor chassis 202 may provide a generic platform that may be dynamically configured to compressively sample a plurality of input signals without requiring any prior knowledge about the input signal or the desired application. Accordingly, the generic nature of the sensor chassis 202 may greatly reduce manufacturing time and complexity. Additionally, the dynamic configuration capability may also enable implementation of a variety of applications using the same sensor chassis 202, thereby reducing deployment costs and efforts. Accordingly, the processing subsystem 214 may analyze the input signal 106 and the corresponding set of parameters to dynamically determine an appropriate CS protocol to achieve the desired sampling performance. Subsequently, the digitizing system 208 may use the determined CS protocol to compressively sample, record and reconstruct the input signal 106. To that end, the digitizing system 208 may include an ADC 220, a clock 222, at least one programmable filter 224 and a recording device 226 for sampling and recording the input signal 106. Therefore, in the present embodiment, the sensor chassis 202 may be equipped to implement a variety of applications without requiring any additional processing devices such as the computing device 112 of
In certain embodiments, the processing subsystem 214 may precondition the input signal 106 to accurately capture salient information corresponding to the input signal 106. By way of example, the salient information may include one or more characteristics corresponding to the input signal 106 such as an input signal structure, an input signal bandwidth, an input signal peak power, and so on. The processing subsystem 214 may use the salient information to introduce sensing diversity to provide a distinct signature or fingerprint to the input signal 106. Moreover, the processing subsystem 214 may analyze the corresponding set of parameters to determine an environmental datum such as ambient noise and sensor chassis characteristics. Particularly, the processing subsystem 214 may evaluate sensor chassis characteristics such as a sampling rate of the ADC 220 and/or a desired sampling performance criterion to determine a CS protocol for sampling the input signal 106 efficiently. In one embodiment, the processing subsystem 214 may query the data repository 122 coupled to the sensor chassis 202 to determine an appropriate CS protocol for compressively sampling the input signal 106. To that end, the data repository 122 may include a plurality of CS protocols devised for different input signals using conventional techniques, such as a distilled sensing technique for astronomical imaging, a non convex compressed sensing for non-Gaussian noise, and so on. Therefore, in accordance with aspects of the present technique, the processing subsystem 214 may query the data repository 122 to determine the CS protocol based on a previously stored correlation, if any, corresponding to the input signal 106 and a CS protocol previously used to compressively sample the input signal 106. In one embodiment, the processing subsystem 214 may select the CS protocol corresponding to the stored correlation to compressively sample the input signal 106. In certain other embodiments, the processing subsystem 214 may further customize the selected CS protocol in accordance with application or user requirements to achieve the desired sampling performance.
Further, the processing subsystem 214 may communicate one or more instructions corresponding to the determined CS protocol to the sampling control unit 210. Subsequently, the sampling control unit 210 may adjust one or more settings corresponding to the programmable filter 224 based on the determined CS protocol to achieve the desired sampling performance. The one or more settings, for example, may correspond to selection of a desired bandwidth to filter out noise, a desired sampling rate, a duty cycle of the input signal 106, a desired sampling accuracy, a number of bits per sample, and so on.
The configuration of the programmable filter 224 may enable the sensor chassis 202 to compressively sample the input signal 106 according to the determined CS protocol. Accordingly, the sensor chassis 202 may employ the ADC 220 and the clock 222 for converting the analog input signal 106 to a sequence of quantized, periodic discrete-time samples. Subsequently, the recording device 226 may record the sampled input signal 106, which may be then be reconstructed by the processing subsystem 214.
Turning to
Determination of a CS protocol generally entails use of salient information such as a set of parameters corresponding to an input signal. As used herein, the term “set of parameters” may refer to a collection of one or more parameters corresponding to the input signal, such as the input signal 106 of
Subsequently, at step 304, the sensor chassis may dynamically determine a CS protocol for compressively sampling the input signal based on the remotely received set of parameters for achieving a desired sampling performance. As previously noted, a processing subsystem, such as the processing subsystem 214 of
In accordance with aspects of the present technique, the processing subsystem may further customize the determined CS protocol to achieve the desired sampling performance. As previously noted, the desired sampling performance may correspond to a maximum acceptable difference between the input signal and a signal reconstructed according to the determined CS protocol. By way of example, in an image compression application, the processing subsystem may determine the CS protocol that not only considers image structure and intra-image correlations, but also adheres to specified error limits during image reconstruction.
Further, in accordance with aspects of the present technique, the processing subsystem may determine the CS protocol to efficiently exploit the structure and other input signal characteristics such as the input signal power spectral density, the input signal average power, and so on. An exemplary implementation of how the processing subsystem may determine the appropriate CS protocol for compressively sampling the input signal will be discussed in greater detail with reference to
In a similar manner, graph 416 is a representation of a total power received at a front end of a receiver, such as the receiver 204, coupled to the sensor chassis. The total power at an instant of time may be defined as a sum of the power spectral density over all frequencies at that instant. The graph 416 indicates that an input signal 418 is only present intermittently over the illustrated period of time. Therefore, while determining the CS protocol, a threshold 420 on the total power may be specified such that compressive sampling may be enabled only when the total power equals or exceeds the threshold 420. Moreover, the depicted input signal characteristics may serve as identifiers of a signal class. Particularly, the input signal characteristics may identify the signal class corresponding to the input signal. Accordingly, in one embodiment, the sensor chassis may query a signal library stored in a data repository coupled to the sensor chassis based on the identified input signal class to determine an appropriate CS protocol for sampling the input signal. As previously noted, the data repository such as the data repository 122 of
With returning reference to
The exemplary method, therefore, describes a technique for dynamically configuring the sensor chassis to compressively sample input signals even where prior information corresponding to the input signals or a desired application is not available. The present technique, thus, allows for fabrication of a generic sensor chassis that may be dynamically configured to implement changing application requirements, thereby reducing the time and complexity involved in setting up and operating CS systems. In accordance with further aspects of the present technique, an alternative embodiment of the exemplary method for compressively sampling the input signal by using a portable sampling control unit is presented and will be discussed in greater detail with reference to
As previously described with reference to the step 302 of
Subsequently, at step 506, the sampling control unit may be programmed to store one or more instructions corresponding to the determined CS protocol. In accordance with aspects of the present technique, the sampling control unit may be an independent unit such as the sampling control unit 128 described with reference to
Thus, in accordance with aspects of the present technique, a generic sensor chassis deployed in a field may remotely receive or detect a set of parameters representative of the input signal to be sampled. An appropriate CS protocol may be determined for sampling the input signal based on the remotely received set of parameters that may include application and user requirements. Instructions corresponding to the determined CS protocol may be stored on a sampling control unit. Subsequently, the sampling control unit having the stored instructions may be installed in the generic sensor chassis. The sampling control unit may, thus, facilitate the generic sensor chassis to compressively sample the input signal according to the determined CS protocol to achieve the desired sampling performance as indicated by step 510. Further, as previously described with reference to step 308 of
The exemplary system and method described hereinabove, thus, enable dynamic configuration of multiple CS protocols to sample a plurality of input signals based on the structure of the input signal, ambient conditions and application and user requirements. The dynamic configuration capability allows quick adaptation to changing application requirements without requiring additional or new hardware, thereby conserving space and battery power. Moreover, the dynamic configuration also allows correction or mitigation of programming errors that may be detected after deployment of the sensor chassis. More particularly, the exemplary method enables fabrication of a generic sensor chassis that may be deployed anywhere and configured ‘on the fly’ to sample a plurality of input signals for a variety of different applications.
While only certain features of the present invention have been illustrated and described herein, many modifications and changes will occur to those skilled in the art. It is, therefore, to be understood that the appended claims are intended to cover all such modifications and changes as fall within the true spirit of the invention.
