Design Of Signal Processing Algorithms Research Articles

DeepRange

Acoustic ranging is a technique for estimating the distance between two objects using acoustic signals, which plays a critical role in many applications, such as motion tracking, gesture/activity recognition, and indoor localization. Although many ranging algorithms have been developed, their performance still degrades significantly under strong noise, interference and hardware limitations. To improve the robustness of the ranging system, in this paper we develop a Deep learning based Ranging system, called DeepRange. We first develop an effective mechanism to generate synthetic training data that captures noise, speaker/mic distortion, and interference in the signals and remove the need of collecting a large volume of training data. We then design a deep range neural network (DRNet) to estimate distance. Our design is inspired by signal processing that ultra-long convolution kernel sizes help to combat the noise and interference. We further apply an ensemble method to enhance the performance. Moreover, we analyze and visualize the network neurons and filters, and identify a few important findings that can be useful for improving the design of signal processing algorithms. Finally, we implement and evaluate DeepRangeusing 11 smartphones with different brands and models, 4 environments (i.e., a lab, a conference room, a corridor, and a cubic area), and 10 users. Our results show that DRNet significantly outperforms existing ranging algorithms.

Improved Auditory-Inspired Signal Processing Algorithm Design for Tracking Multiple Frequency Components

An adaptive synchrony capture filterbank (SCFB) signal processing architecture for tracking signal frequency components was proposed in a previous paper. Detailed analysis and further improvements are presented here. The original SCFB signal processing architecture consisted of a fixed array of traditional, passive linear, gammatone filters in cascade with a bank of three adaptively tunable bandpass filters that form part of a frequency discriminator loop (FDL). The SCFB exhibits many desirable properties for processing speech, music, and other complex sounds. In this article, an improved SCFB architecture is presented along with detailed specification of parameters that yield better frequency tracking. This architecture combines a FDL with a phase-locked loop, a concept borrowed from communication systems, to acquire and track frequency components in a time-varying signal. This modified algorithm estimates frequencies with higher accuracy even in the presence of closely spaced tones in the input. The improved performance of the algorithm is demonstrated by comparing the mean square error in the frequency estimates of synthetic time-varying signals.

Software Simulation Method for Digital if GNSS Signal

Data simulation is an important auxiliary method for baseband signal processing algorithm design of digital IF GNSS receiver. Especially when the high sensitivity and high dynamic receivers are developed, or multipath and cross correlation mitigation techniques are considered, exact and complete tests under all kinds of necessary scenes can be carried out with the IF signal data which is generated by software simulation method. With this method, the developing efficiency can be improved and the cost can be cut down. In this paper, precise model of the IF signal is given firstly through analyzing GNSS receivers RF processing flow. Then, considering the discrete characteristic of software processing, the IF signal simulation flow and block of the simulation software are designed. Finally, the specific design process is given through a design example. When the IF signal simulation software is used for testing baseband signal processing algorithms, GNSS receivers IF signal data under every necessary test scene can be generated just by parameter adjustment.

The Impact of Dynamic Voltage and Frequency Scaling on Multicore DSP Algorithm Design [Exploratory DSP

We connect to the two recent IEEE Signal Processing Magazine special issues on digital signal processing (DSP) on multicore processors (November 2009 and March 2010) and address an issue that was not addressed in the articles there, which we believe has important consequences for DSP algorithm design in the future. The basic observation that we start out with is that in DSP algorithm design, there is very often a tradeoff between the computational effort spent (in terms of the number of operations) and the quality/accuracy of the algorithm output. Herein we discuss the problems that emerge when optimizing algorithms for circuits that support the operation of several parallel cores that operate at different speeds. The understanding, formulation, and solution of these optimization problems require a cross-disciplinary approach that models the interplay between circuits, computer architecture, signal processing, and optimization. While allocation of computational and transmission resources for the purpose of saving energy is an established research topic in other fields (sensor network life time maximization being a notable example), there appears to be relatively little open literature on the type of problems that we discuss here. Taken together, we believe that the challenges we pose are important and that the signal processing algorithm design community is well positioned to tackle them.

A Two-Microphone Noise Reduction System for Cochlear Implant Users with Nearby Microphones—Part I: Signal Processing Algorithm Design and Development

Users of cochlear implant systems, that is, of auditory aids which stimulate the auditory nerve at the cochlea electrically, often complain about poor speech understanding in noisy environments. Despite the proven advantages of multimicrophone directional noise reduction systems for conventional hearing aids, only one major manufacturer has so far implemented such a system in a product, presumably because of the added power consumption and size. We present a physically small (intermicrophone distance 7 mm) and computationally inexpensive adaptive noise reduction system suitable for behind-the-ear cochlear implant speech processors. Supporting algorithms, which allow the adjustment of the opening angle and the maximum noise suppression, are proposed and evaluated. A portable real-time device for test in real acoustic environments is presented.

Snow cover effects on acoustic sensors

Measurements were conducted to determine the effect of a snow cover on acoustic wave propagation during a 50 day period in Alaska. Acoustic waveforms produced by a blank pistol shot were recorded after propagating horizontally over various snow-covered propagation paths and were used to determine the snow cover parameters by comparison with theoretically calculated waveforms. This automatic comparison procedure was successful in determining the average snow cover permeability and depth even when the snow cover depth varied greatly across the propagation path. Although the snow cover properties remained relatively constant during the measurement period, the acoustic measurements were able to determine the changes caused by wind events and new snowfall. Acoustic measurements can provide a rapid, accurate method for determining and monitoring snow cover characteristics. The excess attenuation produced by the snow cover was generally about − 30 dB for 100 m distance and frequencies above 100 Hz. Theoretical calculations also show that even a thin snow cover only 0.02 m thick will affect acoustic pulse propagation. The attenuation and distortion caused by a snow cover can degrade passive acoustic sensor identification and distance estimation, but these detrimental effects can be mitigated by proper design of signal processing algorithms.

Crack Characterization Method with Invariance to Noise Features for Eddy Current Inspection of Fasterner Sites

There is a need for research in eddy current (EC) nondestructive evaluation (NDE) to improve the reliability to detect, locate and size cracks around fastener sites in multi-layer structures while minimizing the overall cost of inspection. The objective of this work is to develop feature extraction and classification algorithms for crack characterization with invariance to noise features for eddy current inspection of fastener sites. Model-based parametric studies were first performed to explore potential features under a wide array of crack, noise and material conditions. Through these studies, several features were identified to have some invariance to the characteristic asymmetric response due to gaps between the fastener and hole, probe liftoff variation, and probe skew. In particular, a promising feature with noise invariance to all non-flaw conditions considered in this study was found through investigating changes in the eddy current response along a circumferential direction in an annulus region away from the hole center. To obtain a measure of this localized crack feature, an approach was developed using a fit of a characteristic function to the data through nonlinear least squares estimation. A model-based optimization approach was also implemented to evaluate the best signal processing algorithm design to distinguish between several classes of crack size. Using this approach, an optimized measure was found to be well correlated with subsurface crack size and insensitive to noise conditions included in this study.

Digital Signal Processing with Interleaved ADC Systems

This paper addresses a problem associated with interleaved ADC systems from the digital signal processing algorithm design perspective. The output streams of an interleaved ADC system are inherently in parallel format. It would be nice if DSP algorithms can be designed to take advantage of the inherently parallel signal streams in the interleaved ADC system without the need of a high speed parallel-to-serial multiplexer. Frequency response of a parallel filter bank is derived. It is found that the overall frequency response is the average of each individual interpolated channel filter plus the aliasing components. The aliasing components come from the deviation of each individual channel from the average response.Results are applied to characterize the gain mismatch of ADC arrays. Sinusoidal response is also investigated. The results can be used to characterize the frequency response mismatch of ADC arrays.

A cross-layer perspective in an uncharted path - Signal processing in random access

In this paper, a cross-layer view for roles of signal processing in random access network and vice versa is presented. The two cases where cross-layer design has a quantifiable impact on system performance are discussed. The first case is a small network (such as wireless LAN) where a few nodes with bursty arrivals communicate with an access point. The design objective is to achieve the highest throughput among users with variable rate and delay constraints. The impact of PHY layer design on MAC protocol is examined and illustrates a tradeoff between allocating resources to the PHY layer and to MAC layer. The second case, in contrast, deals with large-scale sensor networks where each node carries little information but is severely constrained by its computation and communication complexity and most importantly, battery power. This paper emphasizes that the design of signal processing algorithms must take into account the role of MAC and the nature of random arrivals and bursty transmissions.

Structured neural networks for signal classification

Abstract In this paper, artificial neural networks are considered as an emergent alternative to the classical `model-based approach' to the design of signal-processing algorithms. After briefly examining the pros and cons of the neural-network approach, we propose the application of structured neural networks (SNNs) for the classification of signals characterized by different `information sources', such as multisensor signals or signals described by features computed in different domains. The main purpose of such neural networks is to overcome the drawbacks of classical neural classifiers due to the lack of general criteria for `architecture definition' and to the difficulty with interpreting the `network behaviour'. Our structured neural networks are based on multilayer perceptrons with hierarchical sparse architectures that take into account explicitly the `multisource' characteristics of input signals and make it possible to understand and validate the operation of the implemented classification algorithm. In particular, the interpretation of the SNN operation can be used to identify which information sources and which related components are negligible in the classification process. SNNs are compared with both commonly used fully connected multilayer perceptrons and the k-nearest neighbour statistical classifier. Experiments on two multisource data sets related to magnetic-resonance and remote-sensing images are reported and discussed.

Fuzzy time-rank relations and order statistics

The rank ordering of samples is widely used in robust statistics and robust signal processing. Advances in these areas have focused on utilizing joint time-rank (TR) information. The TR information utilized to date is that resulting from a binary, or crisp, relation between the marginal time and rank ordering of samples. This crisp relation, while powerful, contains no information on sample values or spread. This paper generalizes the TR relation through fuzzy set theory. This generalization includes information on sample spread and leads to the concepts of fuzzy TR relations, fuzzy time and rank ordered samples, and fuzzy time and rank indices. These concepts are developed and analyzed through the derivation of fundamental properties. It is shown that the fuzzy TR relations, samples, and indices contain their crisp (standard) counterparts as special cases. These fuzzy generalizations constitute powerful tools that can be exploited in the design of signal processing algorithms.

The past, present, and the future of underwater acoustic signal processing

This is a collection of articles written by members of the Underwater Acoustic Signal Processing (UASP) Technical Committee. The first article, by D. W. Tufts, deals with the history of UASP prior to 1980. In this period, initial mathematical models were developed and the first experimental investigations of underwater acoustic propagation were performed. It was also recognized during this time that there are many similarities between radar and sonar signal processing. The article by J.P. Ianniello deals with research in passive and active sonar from 1980 to the present. Work in this period included experimental verification of algorithms that had been developed in the 1960s and 1970s (e.g. for adaptive beamforming), as well as the development of new approaches, which include acoustic propagation modeling in the design of signal processing algorithms. Such processing is referred to as matched field processing. A common task in passive sonar systems is to estimate the difference in times at which different sensors receive the same signal. Time-delay estimation is a first stage that feeds into subsequent processing blocks. I. Lourtie provides a concise review of work in this field. The article by J.C. Preisig deals with underwater acoustic communications. The underwater channel has several features that make reliable communication a challenging problem. Nevertheless, progress is being made by combining results from ocean acoustic modeling, communication theory, and signal processing. The final article, by J.M.F. Moura, deals with the future of signal processing in the ocean. In addition to considering advances in detection and localization, he deals with new applications such as acoustic tomography, physical oceanography, and synthetic aperture sonar.

Structured neural networks for signal classification

In this paper, artificial neural networks are considered as an emergent alternative to the classical `model-based approach' to the design of signal-processing algorithms. After briefly examining the pros and cons of the neural-network approach, we propose the application of structured neural networks (SNNs) for the classification of signals characterized by different `information sources', such as multisensor signals or signals described by features computed in different domains. The main purpose of such neural networks is to overcome the drawbacks of classical neural classifiers due to the lack of general criteria for `architecture definition' and to the difficulty with interpreting the `network behaviour'. Our structured neural networks are based on multilayer perceptrons with hierarchical sparse architectures that take into account explicitly the `multisource' characteristics of input signals and make it possible to understand and validate the operation of the implemented classification algorithm. In particular, the interpretation of the SNN operation can be used to identify which information sources and which related components are negligible in the classification process. SNNs are compared with both commonly used fully connected multilayer perceptrons and the k-nearest neighbour statistical classifier. Experiments on two multisource data sets related to magnetic-resonance and remote-sensing images are reported and discussed.

Full-spectrum signal processing using a classify-before-detect paradigm

In order to effectively deal with quiet source emissions and elevated ambient noise level in littoral waters, it is important that one understand and exploit the underlying signal microstructure. Experiences indicate that thorough understanding of signal structures is a key to designing robust detection and signal processing algorithms. Therefore, classify-before-detect algorithms are designed and their performance evaluated with passive broadband (PBB) data corrupted by the SWell-EX1 shallow water ambient noise collected near San Diego, CA. The processing strategy is based on (1) exploitation of any microstructure present in target signature by projecting raw data onto appropriate low-dimensional projection spaces, (2) identification of key parameters or ‘‘features’’ crucial in determining the presence of signal, and (3) designing a classifier topology that best matches the underlying feature distribution to minimize modeling errors. Full-spectrum signal processing algorithm design is facilitated by the use of an integrated classification paradigm that takes advantage of an inherent relationship between low-dimensional features and classifier architecture. The analysis results based on the classified PBB signal and the SWell-Ex1 ambient noise indicate that taking advantage of the PBB microstructure improves detection performance by an average of 5 to 15 dB over that of a traditional energy detector.