This paper presents a special matrix factorization based on sparse representation that detects anomalies in video sequences generated with moving cameras. Such representation is made by associating the frames of the target video, that is a sequence to be tested for the presence of anomalies, with the frames of an anomaly-free reference video, which is a previously validated sequence. This factorization is done by a sparse coefficient matrix, and any target-video anomaly is encapsulated into a residue term. In order to cope with camera trepidations, domaintransformations are incorporated into the sparse representation process. Approximations of the transformed-domain optimization problem are introduced to turn it into a feasible iterative process. Results obtained from a comprehensive video database acquired with moving cameras on a visually cluttered environment indicate that the proposed algorithm provides a better geometric registration between reference and target videos, greatly improving the overall performance of the anomaly-detection system.

This paper deals with the challenging problem of visual anomaly detection in a cluttered environment using videos acquired with a moving camera. The anomalies considered are abandoned objects. A new method is proposed for comparing two videos (an anomaly-free reference video and a target one possibly with anomalies) by using convolutional neural networks as feature extractors for a subsequent anomaly-detection stage using a classifier. Two classifier strategies are considered, namely a fully-connected neural network and a random forest algorithm. Results for a comprehensive abandoned object database acquired with a moving camera in a cluttered environment indicate that the proposed architecture can match even the state-of-the-art algorithms in terms of object-detection performance, with a reduction in processing time of 80%.

The detection of abandoned objects in videos from moving cameras is of great importance to automatic surveillance systems that monitor large and visually complex areas. This paper proposes a new method based on sparse decompositions to identify video anomalies associated with abandoned objects. The proposed scheme inherently incorporates synchronization between the reference (anomaly-free) and target (under analysis) sequences thus reducing the implementation complexity of the overall surveillance system. Results indicate that the proposed video-processing scheme can lead to 95% complexity reduction while maintaining excellent detection capability of foreground objects.

This work presents a video-based sequence synchronization algorithm to be used in real-time video surveillance applications. The signals are aligned based on an online dynamic time warping approach that uses only video content information. The algorithm was tested in the alignment of reference and target videos acquired in a cluttered industrial environment with a moving camera. During each recording, the camera follows approximately the same trajectory but its speed along the trajectory may differ among recordings, thus generating time warping between the videos. The results shows that, when considering both alignment error and computational complexity, the use of the mean squared error as similarity metric leads to the best performance among metrics such as the SSIM, absolute error or even sophisticated matching techniques such as histograms of oriented gradients. The proposed method was also tested in a real-time application. Results show that the instantaneous alignment computation of a target frame can be performed successfully, at the cost of a three-fold increase in the error when compared to the case when latency is allowed.

This work presents a novel approach to detect automatically the position of dark fringes in birefringence images. These positions are important for rheology (study of the flow of matter) applications, as they allow indirect measurements of physical properties of molten polymers without having to interact with the material. Our approach uses mathematical morphology techniques to find the patterns that characterize the fringes and detect the center position of each one of them. The results are compared with manual measurements, revealing the superior precision of our method. They have also been applied to the calculation of the stress field of some polymers with good results.

This paper describes an enhancement strategy based on several perceptual-assessment criteria for dereverberation algorithms. The complete procedure is applied to an algorithm for reverberant speech enhancement based on single-channel blind spectral subtraction. This enhancement was implemented by combining different quality measures, namely the so-called QAreverb, the speech-to-reverberation modulation energy ratio (SRMR) and the perceptual evaluation of speech quality (PESQ). Experimental results, using a 4211-signal speech database, indicate that the proposed modifications can improve the word error rate (WER) of speech recognition systems an average of 20%.

We consider the application of sparse-representation and robust-subspace-recovery techniques to detect abandoned objects in a target video acquired with a moving camera. In the proposed framework, the target video is compared to a previously acquired reference video, which is assumed to have no abandoned objects. The detection method explores the low-rank similarities among the reference and target videos, as well as the sparsity of the differences between the two video sequences caused by the unexpected object in the target video. A three-step procedure is then presented adapting a previous low-rank and sparse image representation to the problem at hand. Performance of the proposed technique is verified using a large video database for abandoned-object detection in a cluttered environment. Results demonstrate the technique effectiveness even in the presence of some significant camera shake along its trajectory.

This work presents a framework to be used in the detection of abandoned objects and other video events in a cluttered environment with a moving camera. In the proposed method a target video, that may have features we would like to detect, is compared with a pre-acquired reference video, which is assumed to have no objects nor video events of interest. The comparison is carried out by way of the achieved optimized operators, generated from the reference video, that produce Gaussian outputs when applied to it. Any anomaly of interest in the target video leads to a non-Gaussian output. The method dispenses with the target and reference videos being either synchronized or precisely registered, being robust to rotations and translations between the frames. Experiments show its good performance in the proposed environment.

This paper considers the problem of quantifying the reverberation perception on speech signals. We investigate several combinations of three distinct reverberation-related features (namely, the reverberation time (RT), room spectral variance (RSV), and direct-to-reverberant energy ratio (DRR)), which can be extracted directly from the associated room impulse response. Particular attention is also paid on different post-processing nonlinear mappings in order to provide a more effective quality evaluation algorithm. Results indicate that the RSV feature can be completely disregarded if the RT and DRR estimates are properly weighted. Performance of resulting measure is slightly superior in comparison to previous state-of-the-art method, particularly with respect to the computational cost and robustness (by disregarding the RSV estimation), but also on the statistical correlation level with subjective grades of a large dataset of reverberant speech.

This paper presents three effective proposals for a twostage algorithm for one-microphone reverberant speech enhancement. The original algorithm is divided into two blocks: one that deals with the coloration effect, due to the early reflections, and the other for reducing the long-term reverberation. The proposed modifications consider changing the linearprediction model order, the adaptation stepsize and stop criterion for the first algorithm stage. All the modifications are evaluated by a perceptual-quality measure specific for the speechreverberation context. Experimental results for a 200-signal database show that the proposed improvements yielded an increase of 12% in perceptual measure and a reduction of about 96% in the computation cost when compared to the original framework.

This article presents a low-complexity parametric algorithm for the estimation of the carrier frequency offset (CFO) in orthogonal frequency division multiplexing systems. The proposed algorithm is equivalent to the Unitary ESPRIT algorithm, originally devised for estimating the direction-of-arrival of a wave front, now applied to the CFO scenario. It is verified that the proposed algorithm reduces the overall computational complexity in around 40%, while sustaining the error-rate performance achieved by the standard ESPRIT algorithm.

Prosody transplantation is a speech signal modification procedure usually used to voice transformation or to evaluate the quality of speech synthesizers. In practice, the pitch contour is mapped onto a common segmental content and the target signal is modified adjusting position and length of speech frames to achieve the desired pitch contour and time duration from a speech reference. A new algorithm for prosody transplantation is presented based on a pitch-synchronous feature extraction of the speech signal, unifying the time-aligning and pitch-modification stages. The result is a computationally efficient algorithm for prosody transplantation that maximizes the spectral similarity between the target and reference signals.

This paper revisits the waveform paradigm for coding speech signals, using a multiscale recurrent-pattern matching approach. The so-called MMP (Multidimensional Multiscale Parser) algorithm uses a dictionary which is constantly updated with expansions, contractions, and concatenations of previously encoded segments. This provides a learning ability to the MMP, particularly suited for coding voiced and silent segments of speech. Additional features (nonuniform and auxiliary displacement dictionaries) are considered in order to adjust the MMP learning mechanism for the speech coding problem. Current MMP algorithm achieves a fair-to-good objective score when operating at 8 kbps, as indicated by several speech-coding experiments. This indicates that it may be worthy to further investigate the use of the multiscale recurrent pattern matching paradigm for speech coding.

This paper analyzes the ability of several measurements to quantify the reverberation effect in speech signals. We consider an intrusive scheme, in which the clean and reverberated signals are available, allowing one to estimate the corresponding room impulse response (RIR) signal. An artificial neural network (ANN) is trained for all features and used in a regression approach to estimate the human perceptual evaluation in a mean opinion score (MOS) 1–5 scale. Dimensionality reduction approaches are applied to generate a simpler ANN regression, establishing the most representative features for the problem at hand. A correlation level of 85% with subjective test scores was achieved by reducing the input-vector dimension from 10 to 3, including only the features of reverberation time, room spectral variance, and direct-to-reverberant energy ratio.

This paper describes computationally efficient implementations for the ITU-T G.729 speech codec. Focus is given to the adaptive codebook search, more specifically in the open-loop stage, which first estimates the pitch period of the speech frame being coded. Different strategies are discussed to achieve an excellent compromise between computational complexity and signal quality. The result is an accelerated procedure for the G.729 and G.729A versions, reducing encoding time in about 12% and 9%, respectively, while sustaining original signal quality, as verified by objective and subjective measurements.

In orthogonal frequency division multiplexing (OFDM), carrier frequency offset (CFO) must be mitigated since it generates interference between received symbols transmitted through different sub-carriers. This paper presents a new algorithm for CFO estimation with reduced computational complexity. The new approach is based on the segmentation of the input-signal autocorrelation matrix into the noise and signal subspaces, the latter being employed to estimate the desired CFO. Simulations validate the effectiveness of the new algorithm in comparison to the traditional parametric estimation technique ESPRIT.

This paper constitutes an introduction to the field of quality evaluation of sound (speech and audio) signals. The need for such an assessment is inherent to modern communications: VoIP, mobile phone, or teleconference systems require meaningful measures of performance, which may ultimately assure good service or profitable business. A brief survey on subjective and objective evaluation methods is provided. Recent developments as well as new topics to be investigated are also addressed. Experiments are conducted to illustrate how to validate quality assessment methods.

Parametric methods for direction-of-arrival (DoA) estimation have become very popular due to their low computational complexity and good accuracy. Among parametric DoA algorithms, ESPRIT is one of the most widely used, since it presents low computational complexity in comparison to other parametric methods. Covariance-based DoA (CB-DoA) estimation algorithm provides an even lower complexity alternative to ESPRIT, while imposing the same constraints on the geometry of the receiving array. This letter presents a new algorithm, based on the CB-DoA approach, comprising only real operations. The constraints on the algorithm are the same imposed to the unitary ESPRIT algorithm, allowing a reduction of about 67% on the required computational effort, for equivalent error measure.

Recently, the use of redundant memoryless single-carrier transmitters has been reported as an efficient choice to reduce distortions introduced by finite-impulse response (FIR) channels. In this work, redundant FIR transceivers are proposed to address not only channel frequency selectivity but additive colored noise, which strongly degrades the performance of memoryless transceivers. The transmitter is shown to be paraunitary, resulting in a simple receiver. The proposed system is optimized like a modulated filter bank. Channel shortening and post-combiner equalizers are used to improve system performance. Comparisons with recent proposed schemes are presented, illustrating the efficiency of the new structure for selective channels with colored noise.

Otoacoustic emissions (OAE) are very low-level acoustic responses generated by a healthy cochlea during the normal hearing process. This paper focuses on the detection of click-evoked OAE (CEOAE), using wavelet decomposition, to assist on the diagnosis of possible peripheral hearing disfunctions. A complete wavelet-based modeling procedure for the CEOAE is described, using a powerful toolbox designed for this matter. Simulations indicate that the resulting model reaches correlation levels above 99% with respect to the original CEOAE. Such a match allows an almost-perfect elimination of the stimulus artifact, yielding better detection of the CEOAE and a more precise hearing-system diagnosis.

This paper describes a broad package composed of signal denoising techniques in a uniform framework. The tool aims at simplifying the general tuning of the denoising techniques, by allowing complete visualization of the results and of the parameter influence on the overall performance. The methods implemented include spectral subtraction, Wiener filtering, wavelet transform, and adaptive filtering. Performance evaluation for each method can be assessed through signal-to-noise ratio, crossed correlation, and amplitude reduction criteria. Examples included illustrate the broad applicability for the package tool.

The paper proposes a new transform for musical audio signals. The new transform, so-called modified constant-Q fast filter bank (mCQFFB), yields a log-like description in the frequency domain with improved frequency response when compared to the standard DFT. The improved frequency response of the mCQFFB is achieved by interpreting the sliding-DFT (sDFT) as a multi-stage filter bank, and substituting the sDFT basic filter in this description by a higher-order filter with improved characteristics. The logarithmic spectral scale of the mCQFFB is achieved by the adequate resampling of a chosen channel filter. We also discuss implementation issues related to the mCQFFB and exemplify its application to the analysis of musical signals.

Generation of F0 contours is required for natural-sounding text-to-speech systems. This task can be accomplished using the Fujisaki model, proven to be very good to describe F0 contours based on simple linguistically motivated parameters. However, the extraction of the Fujisaki model parameters is a very intricate problem. Several methods were proposed to solve this problem using iterative optimization techniques. This paper presents a new method capable of extracting the amplitude parameters of the Fujisaki model analytically. The time-marking commands are still obtained via iterative optimization. The result is a more accurate and less computationally intensive amplitude determination due to the proposed closed-form solution. Examples are included illustrating the application of the proposed method.

The frequency-response masking (FRM) method allows the design of selective prototype filters for cosine-modulated filter banks (CMFB) with a reduced number of distinct coefficients. Such a methodology may result in filter banks with a large number of bands (e.g. 1024 or more) and a simplified optimization procedure, as there are less parameters to adjust. This work introduces a numerically efficient optimization procedure, based on a quasi-Newton algorithm, for designing selective FRM-based CMFB. The proposed method uses a perfect-reconstruction FRM prototype filter as a starting point and updates the number of bands of the filter bank during the optimization procedure. Examples provided indicate that figures-of-merit, such as intersymbol and intercarrier interference, for the optimized FRM-CMFB structure are significantly improved without increasing the complexity of the resulting structure.

We use the frequency-response masking (FRM) approach to design prototype filters for cosine-modulated filter banks in the nearly perfect reconstruction case. With such an approach, it is possible to design a FRM filter with overall order almost equal to the direct-form FIR design, with only slight changes in the values of the inter-carrier and inter-symbol interferences and the attenuation of the bank filters. The result is an efficient design with reduced number of multipliers for the overall structure.

A subband analysis of the stationarity characteristics of speech signals is performed. The analysis is based on the evaluation of seven distance measure techniques between consecutive speech segments. The overall ensemble comprised a total of 600 speech sentences each of duration varying from 2 s to 3 s, generated by two male and one female speakers. Segment lengths of 10 ms to 30 ms were considered. The experiments have shown that: most distance measure techniques yielded equivalent results; segments of 10 ms presented greater level of stationary between 0 and 1 kHz; for other lengths, all bands presented similar stationarity degrees. Results can be applied for a proper subband speech processing (e.g., coding) that depends on the stationary assumption of the signals involved.

The composite squared-error (CSE) algorithm results from the combination of the backpropagation algorithm with the pseudo-linear algorithm of Scalero and Tepedelenlioglu (1992). It is observed that the CSE algorithm is able to avoid suboptimal solutions and associated saddle points, thus achieving lower values of the error function than the pseudo-linear algorithm, in fewer iterations than the backpropagation algorithm. In this paper, we investigate the implementation of the CSE algorithm with the concepts of momentum gain, time-varying learning rate, and time-varying combining factor. It is verified that these features can improve the overall properties of the CSE convergence process.

The two versions of the Portuguese language, the one spoken in Portugal and the one spoken in Brazil differ both phonetically and lexically. The paper reports on a menu driven voiced information retrieval system, BomdePapo, which can handle both versions of the language. The system was developed in Delphi during a research project joining teams from the Universidade do Porto, Portugal, and the Universidade Federal do Rio de Janeiro, Brazil. Information retrieval is processed through a tree menu. The system, BomdePapo, is currently being tested on an application for accessing information on sports, culture and politics news. The paper also reports on the main phonetic differences between the two versions.

In the last years, adaptive infinite-duration impulse response filters (IIR) have been studied as a possible alternative to adaptive finite-duration impulse response (FIR) filters. Some of the best known approaches for adaptive IIR filtering include the output error (OE), the equation error (EE), and the Steiglitz-McBride (SM) algorithms. In this paper, a homotopy continuation mapping (HCM) of the SM adaptive algorithm is proposed by using the OE and EE algorithms as basic schemes. The utilization of the homotopy factor results in a simple analysis of the general convergence behavior of the SM algorithm, allowing a direct comparison with the convergence properties of the OE and EE algorithms. Numerical examples are included in order to demonstrate the usefulness of the proposed mapping.>