By Topic

Digital Signal Processing Workshop, 2004 and the 3rd IEEE Signal Processing Education Workshop. 2004 IEEE 11th

Date 1-4 Aug. 2004

Filter Results

Displaying Results 1 - 25 of 91
  • Bayesian estimation of non-stationary AR model parameters via an unknown forgetting factor

    Page(s): 221 - 225
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (382 KB) |  | HTML iconHTML  

    We study Bayesian estimation of the time-varying parameters of a non-stationary AR (autoregressive) process. This is traditionally achieved via exponential forgetting. A numerically tractable solution is available if the forgetting factor is known a priori. This assumption is now relaxed. Instead, we propose joint Bayesian estimation of the AR parameters and the unknown forgetting factor. The posterior distribution is intractable, and is approximated using the variational-Bayes (VB) method. Improved parameter tracking is revealed in simulation. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Cascade classifiers for audio classification

    Page(s): 366 - 370
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (358 KB) |  | HTML iconHTML  

    We present a set of features derived from a model of the early auditory system and the primary auditory cortex. We show that a classification scheme based on AdaBoost works better than a GMM-based method, especially when the feature dimensions are large. Different variations of the AdaBoost-based approach are compared, and it is shown that a cascade of classifiers approach gives high accuracy while reducing the computation time and power by allocating resources proportional to the classification difficulty of the example being considered. For all classifiers considered, both training and testing are performed on one second segments. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Classification of modulation of signals of interest

    Page(s): 226 - 230
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (426 KB) |  | HTML iconHTML  

    In this paper, we present a novel algorithm to automatically characterize and classify the modulation of signals of interest (SOI). To uniquely characterize SOIs that are closely related, we have derived a set of robust features that are based on information theoretic measures such as Renyi entropy and relative entropy and high order statistics. Two measures based on mutual information and relative entropy are also developed to assess the value of a feature - a mutual-information based measure is used to select non-redundant features and a relative entropy based measure is used to select a feature that would improve the classification accuracy. We have developed a multi-class classifier that is constructed by combining a set of binary support vector machines (SVMs). Our experimental results show that the features that we have considered can characterize the modulation of different closely related SOIs very well, and that our modulation classifier is efficient and effective in classifying the modulation of a signal with an average accuracy of 99.5% at SNR of 10 dB. The techniques developed in this paper can be used to classify the modulation of communication signals. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Adaptive refinement in maximally sparse harmonic signal retrieval

    Page(s): 231 - 235
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (388 KB) |  | HTML iconHTML  

    In this paper, we investigate improvements to the iterative re-weighted method of best basis selection for the problem of harmonic retrieval. We describe a computationally efficient (adaptive) refinement of the frequency scale to improve resolution by strategically increasing the dictionary of basis vectors. We also assess the efficacy of using a priori information regarding the number of sinusoidal components, to choose the regularization parameter in the noisy data problem. Computer simulations are used to assess the behavior of various schemes and the following combined approach is suggested: perform adaptive refinement without regularization first before incorporating the regularization. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • The Infinity Project brings DSP brains to robots in the classroom

    Page(s): 88 - 91
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (441 KB) |  | HTML iconHTML  

    The Infinity Project and BEST Robotics, Inc. have teamed up to bring "smarter" robots into high school and college classrooms. Through the use of Hyperception/National Instruments' Visual Application Builder (VAB) and a SPEEDY-33 DSP kit, high school students can rapidly construct and program a robot with capabilities that exceed the radio-controlled robots in after-school competitions today. At the college level, we describe the methods and outcomes of joint robot design projects that include first-year college students taking introductory electrical engineering and mechanical engineering courses at Southern Methodist University. The students formed interdisciplinary design groups to build robots to compete against one another in a "putt-putt" golf competition. These DSP-based prototype robots were developed using the same Infinity Project tools and technology that support future annual BEST Robotics competitions. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • UniCA - a unified classification algorithm for call progress tones

    Page(s): 236 - 240
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (426 KB) |  | HTML iconHTML  

    To increase productivity in an outbound call center, a predictive dialer system is used to automate dialing and to route calls to available agents. Such a system needs to monitor the progress of a call so that the call is sent to an agent depending on how it completes. A call classifier is involved in detecting how a call completes. Typically a call classifier contains various DSP algorithms including ones for classifying different kinds of call progress tones (CPTs). In this paper, we propose a generic algorithm, called UniCA (unified classification algorithm) that can be used for classifying any CPT. In addition, the algorithm is efficient, deals intelligently with ambiguities and noise, and can correctly distinguish between any two theoretically distinguishable CPTs. A version of this algorithm has been implemented and is part of a commercial product. Our implementation was tested with a variety of real and simulated CPTs. The results show that our algorithm performs better than earlier algorithms. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Determining the topology of a telephone system using internally sensed network tomography

    Page(s): 259 - 262
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (313 KB) |  | HTML iconHTML  

    The paper develops methods of characterizing telephone network topology using a small number of internal network sensors. Using methods inspired by medical imaging, telephone call attempts made from many locations around the world provide the measurements. The ability to determine the topology of worldwide telephone networks offers the promise of substantially improving their operating efficiency. This approach can be extended to determine characteristics of data networks as well. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Multi-resolution time-frequency analysis for detection of rhythms of EEG signals

    Page(s): 338 - 341
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (340 KB) |  | HTML iconHTML  

    In recent years, various time-frequency methods have been applied widely For detecting all kinds of feature waves and abnormal waves in EEG signals. But because of their nature and some inherent limitations, their application in EEG analysis has been limited. Considering the excellence and shortcomings of STFT (short time Fourier transform) and wavelets, in the "virtual EEG recording and analysis instrumentation", the multi-resolution time-frequency analysis method, based on STFT and wavelet packet transform, has been introduced to advance the self-adaptive ability for signals, so more flexible division of frequency bands in EEG can be obtained and the basic rhythms in EEG signals can be detected efficiently. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Audio noise suppression based on neuromorphic saliency and phoneme adaptive filtering [speech enhancement]

    Page(s): 361 - 365
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (393 KB) |  | HTML iconHTML  

    An acoustic noise suppression algorithm is described that uses perceptually inspired signal detection techniques to estimate the presence of speech cues in the presence of low SNRs. The signal detector generates frequency-dependent soft-decisions that are used in determining speech presence and in controlling parameters for the speech enhancement gains. With the input of speech segmentation, a phoneme adaptive mechanism is introduced to enhance speech by moderate state-dependent filtering. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Independent component analysis for audio classification

    Page(s): 352 - 355
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (408 KB) |  | HTML iconHTML  

    In this paper, we explore the performance gains achieved by performing independent component analysis (ICA) decomposition on speech features obtained from a model of the early auditory system. ICA projection achieves dimensionality reduction by reducing the redundancy in the feature set (the transformed features are statistically independent). Performance is evaluated for an audio classification environment using a Gaussian mixture model (GMM) classifier and compared against the classification performance of AdaBoost, a wide-margin boosting algorithm. The new features are compared with mel-frequency cepstral coefficients (MFCC) and perceptual linear prediction (PLP) features. We also show that the ICA transformation is well suited for dimensionality reduction of auditory system-inspired features and it significantly improves the classification accuracy. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Accelerating the convergence of POCS algorithms by exponential prediction

    Page(s): 173 - 177
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (359 KB) |  | HTML iconHTML  

    The convergence of projection on convex sets (POCS) algorithms is monotonic and exponential near the point of convergence, so it is reasonable to predict the limit point using a simple exponential regression. For circumstances where the convergence of each coordinate direction is, in fact, monotonic, this results in a significant acceleration of POCS. However, as we show, the convergence in the coordinates is not monotonic at points sufficiently far from the limit point. We develop an algorithm which takes direction changes into account. An example of POCS on bandlimited reconstruction is presented. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Speaker identification in the presence of packet losses

    Page(s): 302 - 306
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (359 KB) |  | HTML iconHTML  

    Gaussian mixture model (GMM)-based speaker identification systems have proved remarkably accurate for large populations using reasonable lengths of high-quality test utterances. Test utterances, however, acquired from cellular telephones or over the Internet (VoIP) may have dropouts due to packet loss. In our research, we have demonstrated that for small packet sizes, these losses can result in degraded accuracy of the speaker identification system. It is shown that by training the GMM model with lossy speech packets, corresponding to the loss rate experienced by the speaker to be identified, significant performance improvement is obtained. In order to avoid the prior estimation of the packet loss rate experienced by the test subject, we propose an algorithm to identify the user based on maximizing the a posteriori probability over the GMM models of the users, trained with several packet loss rates. It is shown that the proposed algorithm provides excellent identification performance. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A row-action alterative to the space-alternating generalized expectation-maximization algorithm for image reconstruction in positron emission tomography

    Page(s): 325 - 328
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (322 KB) |  | HTML iconHTML  

    The space-alternating generalized expectation (SAGE) maximization algorithm has been successfully used in image reconstruction due to its rapid convergence. In this paper, a row-action alternative to the SAGE algorithm (RASAGE) is proposed; it processes the projection data sequentially. In order to speed up the convergence rate, we process the projection lines using a special order in such a way that the sequential projection lines are independent of each other. A relaxation parameter is also used to adjust the projection data update level. Comparison of the RASAGE with SAGE algorithm shows that the former method converges faster than the latter. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Digital signal processing system design: using LabVIEW and TMS320C6000

    Page(s): 10 - 14
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (496 KB) |  | HTML iconHTML  

    LabVIEW, developed by National Instruments, is a graphical programming environment suited for high-level or system-level design. It allows the integration of different signal processing components or subsystems within a graphical framework. This paper introduces a LabVIEW-based book being written for the purpose of furnishing a textbook for DSP laboratory courses offered at many engineering schools. Two examples, namely signal filtering and DTMF receiver, are presented to show how DSP systems are designed using the LabVIEW graphical programming environment. In addition, the book presents the way to implement any desired component(s) of a DSP system on Texas Instruments C6x DSK boards. It is shown how the real-time data exchange feature of the TI Code Composer Studio allows users to establish a communication link between a host PC running LabVIEW and a C6x DSK board. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Multilayer perceptrons applied to entropy-constrained image coding

    Page(s): 129 - 133
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (442 KB) |  | HTML iconHTML  

    To obtain low-complexity solutions for distributed block coding at the focal plane of a camera, we use an entropy-constrained vector quantizer (ECVQ) design algorithm to find a block encoder N that minimizes a rate-distortion Lagrangian cost J over all possible block encoders, and then apply a set of operations to constrain the structure of N to that of a multilayer perceptron (MLP) while keeping J small. The MLP is successful in emulating the ECVQ, with less than 0.2 dB quality loss. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Initialization and bitloading of discrete multi-tone transceivers for channels with unknown impulse response length

    Page(s): 196 - 200
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (380 KB) |  | HTML iconHTML  

    In this paper, we consider a DMT (discrete multi-tone modulation) transceiver with a guard interval of insufficient length. We describe an algorithm that estimates the channel impulse response correctly, independently of its length, while keeping the guard interval at the preset length. The training sequence is designed such that inter-symbol and inter-carrier interference cancel. We then show that calculating the one-tap equalizer per subcarrier, as the inverse of the channel frequency response at the subcarrier frequency, is not optimal in the presence of interference. We also propose an improved bitloading algorithm, based on measured signal-to-noise-and-interference ratios. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Improved lossless audio coding using the noise-shaped IntMDCT

    Page(s): 356 - 360
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (432 KB) |  | HTML iconHTML  

    This paper discusses approximation noise shaping to improve the efficiency of the integer modified discrete cosine transform (IntMDCT)-based lossless audio codec. The scheme is applied to rounding operations associated with lifting steps to shape the noise spectrum towards the low frequency bands. In this paper, constraints on the noise shaping filter and a design procedure with the constraints are discussed. Several noise shaping filters are designed and experimental results showing the improvement are presented. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A neuromorphic IC connection between cortical dendritic processing and HMM classification

    Page(s): 334 - 337
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (372 KB) |  | HTML iconHTML  

    We show connections between dendritic processing structures and hidden-Markov model (HMM) decoding that we arrived at through simultaneous circuit design of these systems. From an integrated circuit (IC) towards a biology perspective, these simple spreading networks relate well to cable theory and are similar to biological structures such as dendrites and cortical cells. Going in the other direction, from IC to classical digital signal processing (DSP), these structure hold similarities to HMM decoders. In order to implement both structures requires a compact array of variable conductance elements. Using floating-gate transistors, we are able to individually vary the conductance of each diffuser element in the array, which dramatically changes the analysis of these arrays. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Evaluation of spherically invariant random process parameters as discriminators for speaker verification

    Page(s): 307 - 310
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (365 KB) |  | HTML iconHTML  

    In this work, we are interested in the potential use of spherically invariant random processes (SIRPs), described by two parameters, for speaker identification. These random processes have been shown to be a more statistically-accurate model for speech than Laplace and Gamma probability density functions. Computation of the two SIRP parameters is fast and simple and storage requirements are obviously small. Although the proposed method does not yield the accuracy of current methods, identification rates are better than random guessing. The work demonstrates the first step for potential use of SIRPs in speaker identification. Usage might include an adjunct role where SIRPs could supplement existing methods to further improve identification or be used to reduce the parameter requirements of existing methods while maintaining accuracy rates. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Design of a flexible and scalable 4×4 MIMO testbed

    Page(s): 178 - 181
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (350 KB) |  | HTML iconHTML  

    The paper reports the development of a flexible and scalable testbed for the implementation and evaluation of signal processing algorithms for multiple antenna (MIMO) systems. At a maximum, four transmit and four receive antennas can be served. Two operational modes are targeted, non real-time block transmission and real-time operation. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Modeling of blood flow velocity and pressure signals using the multipulse method

    Page(s): 320 - 324
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (363 KB) |  | HTML iconHTML  

    The work presented in this paper shows how the multipulse method from digital signal processing can be used to accurately model signals obtained from blood pressure and flow velocity sensors. This model produces very good modelling of the signals on a resolution that allows analysis between heartbeats. The AR coefficients can be transformed to reflection coefficients and tube radii associated with digital wave guides, as well as pole-zero representations. These parameters permit additional insight and interpretation that will produce deeper insight into the biological control mechanisms. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Pitch estimation of polyphony based on controlling delays of comb filters for transcription

    Page(s): 371 - 375
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (326 KB) |  | HTML iconHTML  

    A new pitch estimation method for polyphony based on controlling the delays of comb filters is presented. This method takes the iterative process that extracts a period candidate from the output signal of cascaded comb filters and controls the delay of the comb filter with the extracted period candidate, until the same repetition of a set of extracted candidates is detected. The pitches of polyphony can be estimated from the set of obtained period candidates which makes the average magnitude difference function the minimum. Pitch estimation results for polyphony with this method are presented. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Optoelectronic feedback circuit systems for signal processing of high-resolution images for target tracking

    Page(s): 125 - 128
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (329 KB) |  | HTML iconHTML  

    Techniques for tracking targets in high-resolution images is an important task. However, computations for processing these types of images can be intensive. With recent advancements in LC devices, DSP, and VLSI, alternate methods are being explored. One of these methods is the use of a hybrid optoelectronic architecture. By combining the parallel nature of optics and the processing capabilities of electronics (DSP/VLSI) these system have the potential for providing fast and compact systems for detecting and tracking targets. A unique feature of these optoelectronic systems is a variety of spatial patterns that can be generated, one is the soliton pattern. This soliton pattern can be used to represent the location of a target within an image. In this paper, simulations of the system show trajectories of solitons indicating tracking of the target. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Sampling below the Nyquist rate in interferometric fluorescence microscopy with multi-wavelength measurements to remove aliasing

    Page(s): 329 - 333
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (404 KB) |  | HTML iconHTML  

    A multi-wavelength 3D fluorescence microscope, with transfer functions varying significantly with wavelength, is proposed. This microscope measures multiple wavelengths concurrently and scans through the object at a rate significantly below the Nyquist criterion, which gives a reduced image acquisition time. The sub-Nyquist sampling produces a set of images contaminated by aliasing. Due to the differing transfer functions, the aliasing effects are different in each image. This allows the aliasing operator to be inverted and a single unaliased image to be constructed. This is an application of the generalized sampling expansion first introduced by Papoulis. The instrument is demonstrated through simulation and shown to produce images of a similar quality to those that would be expected from a Nyquist-rate instrument. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Dynamic matched filtering - animating the action [educational aid]

    Page(s): 73 - 78
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (723 KB) |  | HTML iconHTML  

    A Simulink-based block diagram modelling environment is described which makes investigation of demanding DSP concepts such as FIR matched filtering easy and fun. Users interact with the experiment to a remarkable degree, watching scope displays while tuning parameter values or moving sliders to effect model changes during run time in a dynamic fashion. Instrumentation for achieved signal-to-noise ratio sits alongside displays advising the experimenter of theoretically optimal SNR for the current parameter settings. A small example problem using a single-pole noise-shaping filter is seen to be very enlightening, especially since a variable-coefficient matched filter block is employed which is self-designing in response to the prevailing pole radius and resonant frequency selection. View full abstract»

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