By Topic

Digital Signal Processing Workshop and 5th IEEE Signal Processing Education Workshop, 2009. DSP/SPE 2009. IEEE 13th

Date 4-7 Jan. 2009

Filter Results

Displaying Results 1 - 25 of 152
  • [Front cover]

    Page(s): 1
    Save to Project icon | Request Permissions | PDF file iconPDF (338 KB)  
    Freely Available from IEEE
  • [Breaker page]

    Page(s): 1
    Save to Project icon | Request Permissions | PDF file iconPDF (311 KB)  
    Freely Available from IEEE
  • [Breaker page]

    Page(s): ii
    Save to Project icon | Request Permissions | PDF file iconPDF (479 KB)  
    Freely Available from IEEE
  • Contributor listings

    Page(s): iii
    Save to Project icon | Request Permissions | PDF file iconPDF (115 KB)  
    Freely Available from IEEE
  • [Breaker page]

    Page(s): iv
    Save to Project icon | Request Permissions | PDF file iconPDF (6 KB)  
    Freely Available from IEEE
  • Table of contents

    Page(s): v - xviii
    Save to Project icon | Request Permissions | PDF file iconPDF (491 KB)  
    Freely Available from IEEE
  • Haar-Like Filtering with Center-Clipped Emphasis for Speech Detection in Sensornet

    Page(s): 1 - 4
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (4114 KB) |  | HTML iconHTML  

    The use of Haar-like filtering for resourced-constrained speech detection in sensornet application is explored. The simple Haar-like filters having variable filter width and shift width are trained to learn appropriate filter parameters from the training samples to detect speech. To further refine the accuracy, the center-clipped emphasis is proposed as a new degree of freedom for more adaptive Haar-like filter designs. Our method yielded speech/nonspeech classification accuracy of 98.33% for the input length of 0.1 s. Compared with high performance feature extraction method MFCC (mel-frequency cepstrum coefficient), the proposed Haar-like filtering can be approximately 98.40% efficient in terms of the amount of add and multiply computation while capable of achieving the error rate of only 1.63% relative to MFCC. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A Kalman Filter based Fast Noise Suppression Algorithm

    Page(s): 5 - 9
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1557 KB) |  | HTML iconHTML  

    We have proposed a robust noise suppression algorithm with Kalman filter theory [7]. In this paper, we propose a fast noise suppression algorithm by modifying the canonical state space model in [7]. The algorithm aims to achieve robust noise suppression with reduced computational complexity without sacrificing high quality of speech signal. The remarkable features of the proposed algorithm are that it can be realized by 3 multiplications and that it has the same performances or better ones compared with [7] despite the reduction of computational complexity under the same environments, using only the Kalman filter algorithm for the proposed canonical state space model with the colored driving source: (i) a vector state equation is composed of the only speech signal, and (ii) a scalar observation equation is composed of speech signal and additive noise. We have confirmation of validity of the proposed canonical state space model with the colored driving source, and also show the effectiveness through numerical results and subjective evaluation results. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Multiband Excitation for Speech Enhancement

    Page(s): 10 - 15
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (190 KB) |  | HTML iconHTML  

    The speech enhancement algorithm proposed aims to improve the quality of denoised speech by introducing voicing information to a Wiener-type spectral amplitude gain function. A constrained multiband excitation (MBE) model is used to emphasize harmonic components of the glottal input; and a low variance and bias autoregressive multitaper (ARMT) estimate models the vocal tract. Objective and subjective evaluations show improvement over unconstrained models and those using high variance spectrum estimators. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Distributed Audio Coding with Efficient Source Correlation Extraction

    Page(s): 16 - 20
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (266 KB) |  | HTML iconHTML  

    Distributed source coding is one of the enabling technologies for efficient bandwidth utilization in wireless sensor networks and is consequently of great current interest. This paper studies its application to audio signals, using a transform weighted interleaved vector quantization (TWIN-VQ) framework and allowing a sensor node to passively receive and use information from neighboring sensors that is being transmitted to the common joint decoder. Specifically, it uses the linear predictor coefficients generated as side information by TWIN-VQ for one source to determine its frame-by-frame correlations with another source and then conditionally encode MDCT coefficients of the second source. Based on conditional entropy calculations, exploitation of this correlation results in potential improvements of 54-58% in coding efficiency. Using a multi-context adaptive arithmetic coder, an actual bitrate reduction of 38% is achieved. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A Temporally Varying Objective Audio Quality Metric

    Page(s): 21 - 25
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (103 KB) |  | HTML iconHTML  

    Currently, objective audio quality measures designed to mimic the human psychoacoustic system operate by giving a single number to designate the quality of an entire sequence. This methodology is only effective when both the error and the underlying audio sequence are stationary. It is highly desirable to have a quality measure that can predict the quality of time varying errors and sequences. This has applications in many areas, including audio compression, transmission over noisy channels, and further understanding of the human psychoacoustic system. Here we present a metric constructed by adapting several popular audio quality metrics to operate temporally and then combining these metrics into a single objective metric. Results are given over several different error types. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Spectral Multi-Scale Analysis for Multi-Pitch Tracking

    Page(s): 26 - 31
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (4713 KB) |  | HTML iconHTML  

    This paper proposes a robust and accurate multi-pitch estimation method for multiple voices. This method is based on the spectral analysis of the mixture sound multi-scale product. The multi-scale product (PM) consists of making the product of wavelet transform coefficients. The wavelet used is the quadratic spline function. Simulation results showed that the proposed method can robustly estimate FOs for clean speech and for speech mixed with various interferences. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A Low Complexity Noise Suppressor with Hybrid Filterbanks and Adaptive Time-Frequency Tiling

    Page(s): 32 - 36
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (5066 KB) |  | HTML iconHTML  

    This paper proposes a low complexity noise suppressor with hybrid filterbanks and adaptive time-frequency tiling. An analysis hybrid filterbank provides efficient transformation by further decomposing low-frequency bins after a coarse transformation with a short frame size. A synthesis hybrid filterbank also reduces computational complexity in a similar fashion to the analysis hybrid filterbank. Adaptive time-frequency tiling reduces the number of spectral gain calculations. It adaptively generates tiling information in the time-frequency plane based on the signal characteristics. The average number of instructions on a typical DSP chip has been reduced by 30% to 7.5 MIPS in case of mono signals sampled at 44.1 kHz. A Subjective test result shows that the sound quality of the proposed method is comparable to that of the conventional one. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Speaker Identification in Room Reverberation Using GMM-UBM

    Page(s): 37 - 41
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (155 KB) |  | HTML iconHTML  

    Speaker recognition systems tend to degrade if the training and testing conditions differ significantly. Such situations may arise due to the use of different microphones, telephone and mobile handsets or different acoustic conditions. Recently, the effect of the room acoustics on speaker identification (SI) has been investigated and it has been shown that a loss in accuracy results when using clean training and reverberated testing signals. Various techniques like dereverberation, use of multiple microphones, compensations have been proposed to minimize/alleviate the mismatch thereby increasing the SI accuracies. In this paper, we propose to use a Gaussian mixture model-Universal background model (GMM-UBM), with the multiple speaker model approach previously proposed, to compensate for the acoustical mismatch. By using this approach, the SI accuracies have improved over the conventional GMM based SI systems in the presence of room reverberation. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Comparison of Localization Algorithms Using Attenuation Estimates

    Page(s): 42 - 47
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (164 KB) |  | HTML iconHTML  

    In this paper, algorithms for locating sources using attenuation estimates are discussed. These algorithms assume that the sensor locations are known. Six closed form methods and an iterative method are described. A spatial histogram algorithm is introduced. The computational complexities of all the algorithms are discussed and the accuracies of the algorithms are compared as a function of error in the sensor locations. The iterative method based on the GPS technique is the best technique for far field sources. The spatial histogram method is shown to be the most accurate algorithm at the expense of higher computational load, and is the best technique to use for near field sources. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Learning the Intrinsic Dimensions of the Timit Speech Database with Maximum Variance Unfolding

    Page(s): 48 - 53
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (511 KB) |  | HTML iconHTML  

    Modern methods for nonlinear dimensionality reduction have been used extensively in the machine learning community for discovering the intrinsic dimension of several datasets. In this paper we apply one of the most successful ones maximum variance unfolding on a big sample of the well known speech benchmark TIMIT. Although MVU is not generally scalable, we managed to apply to 1 million 39-dimensional points and successfully reduced the dimension down to 15. In this paper we apply some of the state-of-the-art techniques for handling big datasets. The biggest bottleneck is the local neighborhood computation. For 300 K points it took 9 hours while for 1 M points it took 3.5 days. We also demonstrate the weakness of MFCC representation under the k-nearest neighborhood classification since the error rate is more than 50%. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • An Index to Measure "Transient-Ness" of Speech

    Page(s): 54 - 59
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (6591 KB) |  | HTML iconHTML  

    Several studies have shown that emphasis of speech transients can improve the intelligibility of speech in noise. However, each study used a different method to define speech transients, and comparisons across methods are difficult. This paper introduces an index to quantify the transient nature of speech (specifically, the extent to which the onsets and offsets of formants are emphasized compared to steady segments) and applies the index to several versions of transient speech and to processed speech that have been described. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Capon Beamforming for Active Ultrasound Imaging Systems

    Page(s): 60 - 65
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (234 KB) |  | HTML iconHTML  

    Medical ultrasound imaging has unique requirements regarding spatial and amplitude resolution, near-field focusing, wide bandwidth, and real-time operation. Only recently has Capon beamforming been adapted to this. We give examples of images of point targets, cysts, and regions dominated by speckle and discuss how subaperture smoothing, diagonal loading and range/time averaging must be balanced for best performance. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Closed-Form MSE Performance of the Distributed LMS Algorithm

    Page(s): 66 - 71
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (6164 KB) |  | HTML iconHTML  

    Mean-square error (MSE) performance analysis is conducted for a novel distributed least-mean square (D-LMS) algorithm, which is based on consensus, in-network, adaptive estimation using wireless sensor networks (WSNs). For sensor observations that are linearly related to the time-invariant parameter of interest and independent Gaussian data, exact closed-form expressions are derived for the global and sensor-level MSE evolution and steady-state limiting values. Tracking performance is also investigated when the true parameter adheres to a random-walk model. Remarkably for small step-sizes the results accurately extend to the pragmatic setup whereby sensors acquire temporally-correlated (non-)Gaussian data. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A Simplified Predistorter for Distortion Compensation of Parallel Wiener-Type Systems Based on Direct Learning Architecture

    Page(s): 72 - 77
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (146 KB) |  | HTML iconHTML  

    Predistortion of parallel wiener-type systems is considered in this paper. The predistorter is first modeled as a Volterra series. In order to reduce computation complexity, a simplified predistorter model constructed using adaptive FIR filters is proposed. The coefficients of these two suggested predistorters are estimated using the nonlinear filtered-x least mean squares (NFtimesLMS) algorithm based on the direct learning architecture (DLA) approach. The simulation results show that the simplified predistorter can effectively compensate the nonlinear distortion as well as the volterra predistorter with much lower computation complexity. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Stochastic Search Methods to Improve the Convergence of Adaptive Notch Filters

    Page(s): 78 - 83
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (399 KB) |  | HTML iconHTML  

    Adaptive notch filters (ANFs) are known to have convergence problems due to their non-quadratic error surface. We propose two approaches to improve the convergence of the ANF. The first approach is based on the method of stochastic search. The second approach checks to see whether the estimated signal is correlated to the measurement or is just filtered white noise. The ANF is reinitialized when the estimated signal is filtered white noise (i.e. when the ANF misses the right frequency). Both of these methods show superior convergence comparing to the classical Nehorai ANF. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Fundamental Issues in the Stability of Adaptive IIR Filters

    Page(s): 84 - 89
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (125 KB) |  | HTML iconHTML  

    Adaptive IIR filter analysis is more complicated than for the FIR case because (a) some algorithm signals are generated by the adaptive filter itself, and (b) the prediction error relates to the adapted parameters via a filtering operation. Averaging analyses of stability address the first issue by linearization about the convergence point, and the second by using passivity of the error operator. However, published results do not fully account for signal dynamics in the linearization, nor have initial conditions in the passivity analysis been considered. This paper addresses these gaps. Our motivation to revisit these broadly applicable issues is for analyzing recently developed adaptive algorithms that have application to biological systems. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Block Adaptive ICA with a Time Varying Mixing Matrix

    Page(s): 90 - 95
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1027 KB) |  | HTML iconHTML  

    Independent component analysis effectively resolves fixed linear combinations of independent distributions under requirements that include long block length. In environments where the rules of source combination change rapidly, adaptive or block adaptive methods must be deployed; and associated problems of convergence and permutation ambiguity solved. We propose using ICA on overlapping blocks and resolve the permutation ambiguity based on the principle of correlation continuity. We explore the effect of block length, overlap percentage and sufficiency and utility of second order statistics to maintain continuity in the resolved signals. We demonstrate results using simulated test signals and real speech recordings. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Image Interpolation Exploiting Phase Diversity

    Page(s): 96 - 100
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (322 KB) |  | HTML iconHTML  

    In this paper we present an analysis-synthesis system with adaptive synthesis filters. The general formulation can be employed for image compression applications, but can also be used for image size enlargement. The synthesis section is constructed to contain a multiplicity of outputs with different phase characteristics. By exploiting phase diversity in synthesis, improvement in both PSNR and subjective interpolation quality are possible relative to conventional bilinear and bicubic interpolation methods. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Mean and Mean-Square Analysis of the Complex LMS Algorithm for Non-Circular Gaussian Signals

    Page(s): 101 - 106
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (5118 KB) |  | HTML iconHTML  

    The least-mean-square (LMS) algorithm is a useful and popular procedure for adaptive signal processing of both real-valued and complex-valued signals. Past analysis of the complex LMS algorithm has assumed that the input signal vector is circularly-distributed, such that the pseudo-covariance matrix of the input signal is zero. In this paper, we relax this assumption, providing a complete mean and mean-square analysis of the complex LMS algorithm for non-circular Gaussian signals. Our analysis unifies the statistical descriptions of the conventional (real-valued) LMS and complex LMS algorithms as specific cases of our more-general behavioral description, negating the need for a distinction between these two procedures. Simulations indicate that our analysis more-accurately predicts the behavior of complex LMS for non-circular signals as compared to existing analyses in the scientific literature. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.