In this letter, we propose a novel dual-microphone technique for enhancement of speech degraded by background noise and reverberation. Our algorithm is based on a prediction of the coherence function between the noisy input signals, considering both direct and reverberant speech and noise components received by the sensors, and therefore, is capable of dealing with both coherent and diffuse noise. After predicting the coherence function, the signal to noise ratio (SNR) can be estimated by solving a quadratic equation obtained from the real and imaginary parts of the function. Objective evaluation in a room with reverberation time T60= 465 ms, demonstrated noticeable improvements in SNR and quality of the outputs processed with the proposed algorithm over the baseline (front microphone), as well as a recently proposed coherence-based noise reduction algorithm.

Communication system mismatch represents a major influence for loss in speaker recognition performance. While microphone and handset differences have been considered in the NIST SRE, nonlinear communication system differences, such as modulation/demodulation (Mod/DeMod) carrier drift, have yet to be considered. In this study, an algorithm for estimating and correcting Mod/DeMod frequency offsets distortion in signal sideband modulation (SSB) speech is formulated based on two processing steps. In the first step, the offset of speech can be roughly scaled to a small frequency interval, which eliminates the ambiguity caused by periodicity of the spectrum. The second step performs fine-tuning within the pre-determined interval. For the first time, a statistical framework is developed for unique interval detection, where an innovative acoustic feature is proposed to represent different offsets and state-of-the-art techniques, the total variety method and PLDA, are applied. Speaker recognition experiments on SSB speech obtained from DAPPA RATS corpus show that a significant performance improvement (up to 50% relative improvement in EER) for speaker verification in SSB speech can be obtained by the proposed estimation and compensation method.

This study proposes a speech enhancement algorithm to improve speech intelligibility for hearing impaired listeners in adverse conditions. The proposed algorithm is based on a long term harmonic model, where the harmonics of target speech are more distinguished from noise spectrum interference. Our method consists of two stages: i) Prominent pitch estimation based on long term harmonic feature analysis and neural network classification. ii) Target speech spectrum estimation with pitch information based on long term noise spectrum extraction. The listening experiment with EAS vocoder speech shows that our algorithm is substantially beneficial for cochlear implant recipients to perceive speech in noisy environment in terms of word recognition rate.

Mask-based objective speech-intelligibility measures have been successfully proposed for evaluating the performance of binary masking algorithms. These objective measures were computed directly by comparing the estimated binary mask against the ground truth ideal binary mask (IdBM). Most of these objective measures, however, assign equal weight to all time-frequency (T-F) units. In this study, we propose to improve the existing mask-based objective measures by weighting each T-F unit according to its target or masker loudness. The proposed objective measure shows significantly better performance than two other existing mask-based objective measures.

This paper introduces a novel technique for signal-to-noise ratio (SNR) estimation for scenarios where two closely-spaced microphones are available. The proposed technique utilizes the real and imaginary parts of the coherence function between the input signals to estimate the SNR without assuming prior knowledge of the noise statistics. The corresponding dual-microphone speech enhancement algorithm utilizes a Wiener filter as a gain function constructed using the SNR values computed by the coherence function. Since the proposed SNR estimation technique does not require access to noise statistics, it can be applied in situations where interfering speakers are present. An adaptive speech reception threshold (SRT) test was used to assess the intelligibility of speech processed by the proposed algorithm in scenarios where one or two interfering talkers were present in anechoic and reverberant conditions. Intelligibility listening tests were conducted with both normal-hearing (NH) and cochlear implant (CI) listeners. Results revealed significant improvements in intelligibility and quality over a (baseline) fixed directional algorithm and a well-established beamformer algorithm. In a nearly anechoic room with competing talkers, the improvement in SRT obtained relative to the directional microphone ranged from 5-10 dB, while the improvement obtained by the beamformer was about 2 dB. In reverberant environments, the improvement in SRT remained high (4-7 dB) atT60= 220 ms, and decreased to 1-2 dB atT60= 465 ms. Overall, the proposed algorithm provided significant benefits in intelligibility in anechoic and mildly reverberant environments making it suitable for hearing aid and cochlear implant applications.

A blind (non-ideal) time-frequency (T-F) masking technique is proposed for suppressing reverberation. A binary mask is estimated at each T-F unit by extracting a single variance-based feature from the reverberant signal and comparing its value against an adaptive threshold. The performance of the estimated binary mask is evaluated using intelligibility listening tests with hearing impaired listeners in four moderate to highly reverberant conditions. Results indicated that the proposed T-F masking technique yielded significant improvements in intelligibility even in highly reverberant conditions (T60 = 1.0 s). This improvement was attributed to the recovery of the vowel/consonant boundaries which are severely smeared in reverberation.

In this paper, we evaluate the performance of six intrusive ob-jective measures as intelligibility predictors of degraded speech for cochlear implant (CI) users. Three practical environmental degradation scenarios are considered: reverberation alone, ad-ditive noise alone, and noise-plus-reverberation. A subjective intelligibility test was performed with eleven cochlear implant users and objective measures were evaluated using four per-formance metrics: Pearson, Spearman rank, and sigmoid-fitted correlation coefficients, and the root mean square error. It was observed that existing metrics performed well in the noise-alone scenarios, but obtained lower performance in the reverberation-alone scenario and in many cases, unacceptable results in the noise-plus-reverberation scenario. It is concluded that further work is still needed in order to accurately predict speech intel-ligibility ratings for CI users, particularly in environments cor-rupted by reverberation.

A novel dual-microphone speech enhancement technique is proposed in the present paper. The technique utilizes the coherence between the target and noise signals as a criterion for noise reduction and can be generally applied to arrays with closely-spaced microphones, where noise captured by the sensors is highly correlated. The proposed algorithm is simple to implement and requires no estimation of noise statistics. In addition, it offers the capability of coping with multiple interfering sources that might be located at different azimuths. The proposed algorithm was evaluated with normal hearing listeners using intelligibility listening tests and compared against a well-established beamforming algorithm. Results indicated large gains in speech intelligibility relative to the baseline (front microphone) algorithm in both single and multiple-noise source scenarios. The proposed algorithm was found to yield substantially higher intelligibility than that obtained by the beamforming algorithm, particularly when multiple noise sources or competing talker(s) were present. Objective quality evaluation of the proposed algorithm also indicated significant quality improvement over that obtained by the beamforming algorithm. The intelligibility and quality benefits observed with the proposed coherence-based algorithm make it a viable candidate for hearing aid and cochlear implant devices.

In this paper, we evaluate the performance of existing and new objective measures in terms of predicting the quality of reverberant speech and speech enhanced by dereverberation algorithms. We use subjective quality ratings designed to evaluate the quality of speech along three dimensions: speech coloration, reverberation tail effect and overall speech quality. Experimental results assess the correlations between the proposed objective quality measures and the three subjective rating scales and suggest that PESQ-based measures can very reliably predict the quality of reverberant and dereverberated speech.

Existing speech enhancement algorithms can improve speech quality but not speech intelligibility, and the reasons for that are unclear. In the present paper, we present a theoretical framework that can be used to analyze potential factors that can influence the intelligibility of processed speech. More specifically, this framework focuses on the fine-grain analysis of the distortions introduced by speech enhancement algorithms. It is hypothesized that if these distortions are properly controlled, then large gains in intelligibility can be achieved. To test this hypothesis, intelligibility tests are conducted with human listeners in which we present processed speech with controlled speech distortions. The aim of these tests is to assess the perceptual effect of the various distortions that can be introduced by speech enhancement algorithms on speech intelligibility. Results with three different enhancement algorithms indicated that certain distortions are more detrimental to speech intelligibility degradation than others. When these distortions were properly controlled, however, large gains in intelligibility were obtained by human listeners, even by spectral-subtractive algorithms which are known to degrade speech quality and intelligibility.

Statistical estimators of the magnitude-squared spectrum are derived based on the assumption that the magnitude-squared spectrum of the noisy speech signal can be computed as the sum of the (clean) signal and noise magnitude-squared spectra. Maximum a posterior (MAP) and minimum mean square error (MMSE) estimators are derived based on a Gaussian statistical model. The gain function of the MAP estimator was found to be identical to the gain function used in the ideal binary mask (IdBM) that is widely used in computational auditory scene analysis (CASA). As such, it was binary and assumed the value of 1 if the local SNR exceeded 0 dB, and assumed the value of 0 otherwise. By modeling the local instantaneous SNR as an F-distributed random variable, soft masking methods were derived incorporating SNR uncertainty. The soft masking method, in particular, which weighted the noisy magnitude-squared spectrum by the a priori probability that the local SNR exceeds 0 dB was shown to be identical to the Wiener gain function. Results indicated that the proposed estimators yielded significantly better speech quality than the conventional MMSE spectral power estimators, in terms of yielding lower residual noise and lower speech distortion.

This paper introduces a new speech enhancement approach based on the design of a frequency-specific composite gain function for Wiener filtering. Motivated by the recently established finding that the acoustic cues at low frequencies can improve speech recognition in noise by the combined electric and acoustic stimulation technique, a less aggressive gain function is applied to replace the Wiener filter's conventional rigid gain function at low frequencies. With this modification, the proposed approach is able to recover more low frequency (LF) components and enhance the speech quality in noisy environments. An adaptive procedure is utilized to determine the low frequency boundary. Objective evaluation, based on the PESQ measure, revealed that the proposed approach with adaptive LF boundary improved significantly the PESQ scores compared to the scores obtained with the conventional Wiener filter in steady-state noise, white noise and babble interference conditions. Speech quality was also found to be significantly enhanced in the car noise conditions when the less aggressive gain function was selectively applied only to the consonants.

This paper presents the real-time implementation of wavelet-based Advanced Combination Encoder on PDA platforms for cochlear implant studies. Three real-time implementations using the conventional FFT, our previous recursive DFT, and the wavelet transform are compared in terms of computational complexity, processing time and fixed-point accuracy. The results obtained show that this new real-time implementation on PDA platforms is computationally comparable to our previous real-time recursive DFT implementation while achieving a higher accuracy or a lower quantization error.

In this work, we develop a two-stage blind dereverberation scheme for reverberant speech enhancement. The proposed algorithm operates by first splitting the reverberant inputs into different subbands. In the first stage, the inverse filters are estimated using the blind multiple input-output inverse-filtering theorem (MINT), while in the second stage our approach suppresses late reverberant energy by subtracting the power spectrum of the late impulse components from the power spectrum of the inverse filtered speech signal. Experimental results with reverberant speech indicate that the performance of the proposed algorithm is substantially better when compared to existing dereverberation algorithms.

While most speech enhancement algorithms improve speech quality, they do not improve speech intelligibility in noise. The reasons for that remain unclear. In this paper, we present a theoretical framework that can be used to analyze potential factors influencing the intelligibility of processed speech. It is hypothesized that if distortions are properly controlled, then large gains in intelligibility can be achieved. To assess the perceptual effect of the various distortions that can be introduced by speech enhancement algorithms, intelligibility tests are conducted with human listeners. The results indicated that certain distortions are more critical than others. The result of listening tests suggested that when these distortions are properly controlled, substantial gains in intelligibility can be obtained.

This paper presents a novel speech enhancement algorithm that can substantially improve the signal-to-residual spectrum ratio by combining statistical estimators of the spectral magnitude of the speech and noise. The noise spectral magnitude estimator is derived from the speech magnitude estimator, by appropriately transforming the a priori and the a posteriori SNR values. By expressing the signal-to-residual spectrum ratio as a function of the estimator's gain function, we derive a hybrid strategy that can improve the signal-to-residual spectrum ratio when the a priori and the a posteriori SNR are detected to be lower than 0 dB. Experimental results showed that the signal-to-residual spectrum ratio as well as the PESQ scores can be improved substantially in stationary and quasi-stationary noise conditions with the proposed hybrid estimators. Informal listening tests revealed improved speech quality and no musical noise.

It has been shown that large gains in speech intelligibility can be obtained by using the binary mask approach which retains the time-frequency (T-F) units of the mixture signal that are stronger than the interfering noise (masker) (i.e., SNR>0 dB), and removes the T-F units where the interfering noise dominates. In this paper, we introduce a new binary mask for improving speech intelligibility based on noise distortion constraints. A binary mask is designed to retain noise overestimated T-F units while discarding noise underestimated T-F units. Listening tests were conducted to evaluate the new binary mask in terms of intelligibility. Results from the listening tests indicated that large gains in intelligibility can be achieved by the application of the proposed binary mask to noise-corrupted speech even at extremely low SNR levels (-10 dB). Index Terms: speech intelligibility, noise estimation, speech enhancement

A new binary mask is introduced for improving speech intelligibility based on magnitude spectrum constraints. The proposed binary mask is designed to retain time-frequency (T-F) units of the mixture signal satisfying a magnitude constraint while discarding T-F units violating the constraint. Motivated by prior intelligibility studies of speech synthesized using the ideal binary mask, an algorithm is proposed that decomposes the input signal into T-F units and makes binary decisions, based on a Bayesian classifier, as to whether each T-F unit satisfies the magnitude constraint or not. Speech corrupted at low signal-to-noise (SNR) levels (-5 and 0 dB) using different types of maskers is synthesized by this algorithm and presented to normal-hearing listeners for identification. Results indicated substantial improvements in intelligibility over that attained by human listeners with unprocessed stimuli.

While most speech enhancement algorithms improve speech quality, they may not improve speech intelligibility in noise. This paper focuses on the development of an algorithm that can be optimized for a specific acoustic environment and improve speech intelligibility. The proposed method decomposes the input signal into time-frequency (T-F) units and makes binary decisions, based on a Bayesian classifier, as to whether each T-F unit is dominated by the target signal or the noise masker. Target-dominated T-F units are retained while masker-dominated T-F units are discarded. The Bayesian classifier is trained for each acoustic environment using an incremental approach that continuously updates the model parameters as more data become available. Listening experiments were conducted to assess the intelligibility of speech synthesized using the incrementally adapted models as a function of the number of training sentences. Results indicated substantial improvements in intelligibility (over 60% in babble at -5 dB SNR) with as few as ten training sentences in babble and at least 80 sentences in other noisy conditions.

This paper presents a real-time and interactive implementation of the recursive Fourier transform approach on PDA platforms for cochlear implant signal processing applications. PDA platforms provide a cost-effective and portable platform for cochlear implant studies. The computational complexities of the two commonly used signal processing strategies that are being used in commercial cochlear implants are compared. A more computationally efficient approach using recursive Fourier transform is discussed and a real-time implementation of this approach is then accomplished on a PDA platform. Different versions of the implementation are examined and compared in terms of speed and accuracy.

In speech enhancement applications, a validated metric of noise reduction performance is vital in the relative ranking of noise reduction algorithms and in enhancing the performance of a noise reduction algorithm. subjective scores of enhanced speech remain the yardstick for performance, but objective metrics that emulate subjective evaluations are preferred for cost- and time-effectiveness. In this paper, we analyze the performance of two objective methods for predicting the quality of enhanced speech. The first method employs the coherence-based speech intelligibility index, while the second method uses features derived from the Moore-Glasberg auditory model. In both cases, the features are mapped to a quality score using the Bayesian modeling approach. Results show that the combination of the auditory model-based feature set and the Bayesian modeling provides the best performance in predicting the quality scores of enhanced speech.

The ideal binary mask, often used in robust speech recognition applications, requires an estimate of the local SNR in each timefrequency (T-F) unit. A data-driven approach is proposed for estimating the instantaneous SNR of each T-F unit. By assuming that the a priori SNR and a posteriori SNR are uniformly distributed within a small region, the instantaneous SNR is estimated by minimizing the localized Bayes risk. The binary mask estimator derived by the proposed approach is evaluated in terms of hit and false alarm rates. Compared to the binary mask estimator that uses the decision-directed approach to compute the SNR, the proposed data-driven approach yielded substantial improvements (up to 40%) in classification performance, when assessed in terms of a sensitivity metric which is based on the difference between the hit and false alarm rates. Index Terms: ideal binary mask, SNR estimation, Bayes risk 1.

In this letter we propose a novel approach for two-microphone enhancement of speech corrupted by reverberation. Our approach steers computational resources to filter coefficients having the largest impact on the error surface and therefore only updates a subset of coefficients in every iteration. Experimental results carried out in a realistically reverberant setup indicate that the performance of the proposed algorithm is comparable to the performance of its full-update counterpart.

Subband adaptive filtering (SAF) techniques play a prominent role in designing active noise control (ANC) systems. They reduce the computational complexity of ANC algorithms, particularly, when the acoustic noise is a broadband signal and the system models have long impulse responses. In the commonly used uniform-discrete Fourier transform (DFT)-modulated (UDFTM) filter banks, increasing the number of subbands decreases the computational burden but can introduce excessive distortion, degrading performance of the ANC system. In this paper, we propose a new UDFTM-based adaptive subband filtering method that alleviates the degrading effects of the delay and side-lobe distortion introduced by the prototype filter on the system performance. The delay in filter bank is reduced by prototype filter design and the side-lobe distortion is compensated for by oversampling and appropriate stacking of subband weights. Experimental results show the improvement of performance and computational complexity of the proposed method in comparison to two commonly used subband and block adaptive filtering algorithms.

The intima-media thickness (IMT) of the common carotid artery (CCA) is widely used as an early indicator of cardiovascular disease (CVD). It was proposed but not thoroughly investigated that the media layer (ML), its composition and texture, may be indicative for identifying the risk of stroke and differentiating between patients of high and low risk. In this study we investigate the usefulness of texture analysis of the ML of the CCA. The study was performed on 100 longitudinal ultrasound images acquired from asymptomatic subjects at risk of atherosclerosis. The images were separated into three different age groups, namely below 50, 50 to 60, and above 60 years old. A total of 61 different texture features were extracted from the intima-media complex (IMC), ML and the intima layer (IL). The IMC and ML were segmented manually by a neurovascular expert and automatically by a snakes segmentation system. It was shown that texture features extracted from the IL, ML and IMC are significantly different (mean, gray scale median (GSM), standard deviation, contrast, difference variance, periodicity) and that some of them can be associated with the increase (difference variance, entropy) or decrease (GSM) of patientpsilas age. It was also shown that the GSM of the ML falls linearly with increasing ML thickness (MLT) and with increasing age. Further research on more subjects is required for estimating other features that may provide information for patients at risk of stroke.

Signals and systems lab courses that are currently offered at many universities are mostly based on text-based programming languages and environments, in particular MATLAB. This paper presents an alternative programming approach in these courses by combining textual with graphical programming. This hybrid programming approach offers code interactivity in a time-efficient manner. Example labs are presented to demonstrate how this approach allows students to interact and experiment with their codes to gain a better understanding of the signal processing concepts.

In this paper, we evaluate the performance of several objective measures in terms of predicting the quality of noisy speech enhanced by noise suppression algorithms. The objective measures considered a wide range of distortions introduced by four types of real-world noise at two signal-to-noise ratio levels by four classes of speech enhancement algorithms: spectral subtractive, subspace, statistical-model based, and Wiener algorithms. The subjective quality ratings were obtained using the ITU-T P.835 methodology designed to evaluate the quality of enhanced speech along three dimensions: signal distortion, noise distortion, and overall quality. This paper reports on the evaluation of correlations of several objective measures with these three subjective rating scales. Several new composite objective measures are also proposed by combining the individual objective measures using nonparametric and parametric regression analysis techniques.

In this paper, we report on the evaluation of intelligibility of speech enhancement algorithms. IEEE sentences were corrupted by four types of noise including babble, car, street and train at two SNR levels (0 dB and 5 dB), and then processed by eight speech enhancement methods encompassing four classes of algorithms: spectral subtractive, subspace, statistical model based and Wiener-type algorithms. The processed speech files were presented to normal hearing listeners for identification in formal listening tests. Intelligibility was assessed as the percentage of words identified correctly. This paper reports the results of the intelligibility tests.

This paper describes a highly practical blind signal separation (BSS) scheme operating on subband domain data to blindly segregate convolutive mixtures of speech. The proposed method relies on spatiotemporal separation carried out in the time domain by using a multichannel blind deconvolution (MBD) algorithm that enforces separation by entropy maximization through the popular natural gradient algorithm (NGA). Numerical experiments with binaural impulse responses affirm the validity and illustrate the practical appeal of the presented technique even for difficult speech separation setups.

This paper presents the real-time implementation of a cochlear implant signal processing system on PDA platforms. PDAs are chosen as they provide portable and cost-effective computation platforms. To gain software flexibility and interactivity, the LabVIEW graphical programming environment is used. The paper discusses the optimization steps which are taken to achieve a real-time throughput. These steps consist of using dynamic link libraries, utilizing efficient memory allocation, and performing fixed-point arithmetic. These steps are general purpose in the sense that the same steps can be deployed for real-time implementation of other clinical or industrial signal processing applications on PDAs.

In this paper, we propose and evaluate an integrated system for the segmentation of atherosclerotic plaque in ultrasound imaging of the carotid artery based on normalization, speckle reduction filtering, and four different snakes segmentation methods. These methods are the Williams and Shah, Balloon, Lai and Chin, and the gradient vector flow (GVF) snake. The performance of the four different plaque snakes segmentation methods was tested on 80 longitudinal ultrasound images of the carotid artery using receiver operating characteristic (ROC) analysis and the manual delineations of an expert. All four methods were very satisfactory and similar in all measures evaluated, with no significant differences between them; however, the Lai and Chin snakes segmentation method gave slightly better results. Concluding, it is proposed that the integrated system investigated in this study could be used successfully for the automated segmentation of the carotid plaque.

We report on the development of a noisy speech corpus suitable for evaluation of speech enhancement algorithms. This corpus is used for the subjective evaluation of 13 speech enhancement methods encompassing four classes of algorithms: spectral subtractive, subspace, statistical-model based and Wiener algorithms. The subjective evaluation was performed by Dynastat, Inc. using the ITU-T P.835 methodology designed to evaluate the speech quality along three dimensions: signal distortion, noise distortion and overall quality. This paper reports the results of the subjective tests

We present an adaptive enhancement method for speech signal corrupted by the 3 Tesla (3T) functional Magnetic Resonance Imaging (fMRI) acoustic noise. The noisy speech signal is filtered into several frequency bands and noise cancellation is applied to each band independently. We use Normalized Least Mean Squares (NLMS) algorithm for updating the co-efficients of the adaptive filters in both wideband and in multi-band implementations. The NLMS algorithm is used to ensure faster convergence and stability. The multi-band method shows a better performance than the conventional wideband method resulting in a better noise reduction. It is shown that the distortion introduced by the band-pass filters limits the overall speech enhancement when the number of bands increases beyond an optimal value.

In this paper, we evaluate the performance of several objective measures in terms of predicting the quality of noisy speech enhanced by noise suppression algorithms. The objective measures considered a wide range of distortions introduced by four types of real-world noise at two SNRs by four classes of speech enhancement algorithms: spectral subtractive, subspace, statistical-model based and Wiener algorithms. The subjective quality ratings were obtained using the ITU-T P.835 methodology designed to evaluate the speech quality along three dimensions: signal distortion, noise distortion and overall quality. This paper reports the correlations of five common objective measures with these three subjective measures. Improvements to the PESQ measure are reported along with new composite objective measures. Index Terms: speech enhancement, noise reduction, ITU-T P.835, objective measures, subjective listening test, correlation analysis.

This paper focuses on optimal estimators of the magnitude spectrum for speech enhancement. We present an analytical solution for estimating in the MMSE sense the magnitude spectrum when the clean speech DFT coefficients are modeled by a Laplacian distribution and the noise DFT coefficients are modeled by a Gaussian distribution. Furthermore, we derive the MMSE estimator under speech presence uncertainty and a Laplacian model. Results indicated that the Laplacian based MMSE estimator yielded less residual noise in the enhanced speech than the traditional Gaussian-based MMSE estimator.

The traditional minimum mean-square error (MMSE) estimator of the short-time spectral amplitude is based on the minimization of the Bayesian squared-error cost function. The squared-error cost function, however, is not subjectively meaningful in that it does not necessarily produce estimators that emphasize spectral peak (formants) information or estimators which take into account auditory masking effects. To overcome the shortcomings of the MMSE estimator, we propose in this paper Bayesian estimators of the short-time spectral magnitude of speech based on perceptually motivated cost functions. In particular, we use variants of speech distortion measures, such as the Itakura-Saito and weighted likelihood-ratio distortion measures, which have been used successfully in speech recognition. Three classes of Bayesian estimators of the speech magnitude spectrum are derived. The first class of estimators emphasizes spectral peak information, the second class uses a weighted-Euclidean cost function that implicitly takes into account auditory masking effects, and the third class of estimators is designed to penalize spectral attenuation. Of the three classes of Bayesian estimators, the estimators that implicitly take into account auditory masking effect performed the best in terms of having less residual noise and better speech quality.

This paper addresses the problem of formant frequency estimation of speech signals corrupted by colored noise. The spectrum is sequentially segmented into K segments so that each segment contains a single formant. A segmentation metric based on Wiener filter theory is proposed for determining the segment boundaries. A peak-picking algorithm is used for estimating the formant frequencies in each segment. Results obtained using vowels embedded in +5 dB S/N speech-shaped noise, indicated that the proposed algorithm produced formant frequencies which were comparable to those estimated in quiet.

A noise estimation algorithm is proposed for highly nonstationary noise environments. The noise estimate is updated by averaging the noisy speech power spectrum using a time and frequency dependent smoothing factor, which is adjusted based on signal presence probability in subbands. Signal presence is determined by computing the ratio of the noisy speech power spectrum to its local minimum, which is computed by averaging past values of the noisy speech power spectra with a look-ahead factor. The local minimum estimation algorithm adapts very quickly to highly non-stationary noise environments. This was confirmed with formal listening tests that indicated that our noise estimation algorithm when integrated in speech enhancement was preferred over other noise estimation algorithms.

A frequency domain optimal linear estimator is proposed which incorporates the masking properties of the human auditory system to make the residual noise distortion inaudible. The use of wavelet-thresholded multitaper spectra is also proposed for frequency-domain speech enhancement methods as an alternative to the traditional fast Fourier transform (FFT)-based magnitude spectra. Experiments with multitalker babble noise indicated that the proposed estimator outperformed the minimum mean-square error log-spectral amplitude estimator (MMSE-LSA), particularly when wavelet-thresholded multitaper spectra were used in place of the FFT spectra.

It is well known that the "musical noise" encountered in most frequency domain speech enhancement algorithms is partially due to the large variance estimates of the spectra. To address this issue, we propose in this paper the use of low-variance spectral estimators based on wavelet thresholding the multitaper spectra for speech enhancement. A short-time spectral amplitude estimator is derived which incorporates the wavelet-thresholded multitaper spectra. Listening tests showed that the use of multitaper spectrum estimation combined with wavelet thresholding suppressed the musical noise and yielded better quality than the subspace and MMSE algorithms.

A new algorithm is developed for voiced-unvoiced speech discrimination in noise. Short segments of speech are modeled as a sum of basis functions from a Gabor dictionary. In each iteration, a Gabor atom is fitted (using the matching pursuit algorithm) to the residual obtained by subtracting the best-fit Gabor atom from the previous residual. Multiple discriminant analysis is used to reduce the dimensionality of the vector of Gabor coefficients to give a low-dimensional feature vector for classification. A radial basis function neural network is trained on the reduced feature vector set to discriminate between voiced and unvoiced speech/silence segments. On a database of 62 sentences in 5-dB SNR speech-shaped noise, 84% correct classification accuracy was obtained.

A generalized subspace approach is proposed for enhancement of speech corrupted by colored noise. A nonunitary transform, based on the simultaneous diagonalization of the clean speech and noise covariance matrices, is used to project the noisy signal onto a signal-plus-noise subspace and a noise subspace. The clean signal is estimated by nulling the signal components in the noise subspace and retaining the components in the signal subspace. The applied transform has built-in prewhitening and can therefore be used in general for colored noise. The proposed approach is shown to be a generalization of the approach proposed by Y. Ephraim and H.L. Van Trees (see ibid., vol.3, p.251-66, 1995) for white noise. Two estimators are derived based on the nonunitary transform, one based on time-domain constraints and one based on spectral domain constraints. Objective and subjective measures demonstrate improvements over other subspace-based methods when tested with TIMIT sentences corrupted with speech-shaped noise and multi-talker babble.

A new perceptually motivated approach is proposed for enhancement of speech corrupted by colored noise. The proposed approach takes into account the frequency masking properties of the human auditory system and reduces the perceptual effect of the residual noise. This new perceptual method is incorporated into a frequency-domain speech enhancement method and a subspace-based speech enhancement method. A better power spectrum/autocorrelation function estimator was also developed to improve the performance of the proposed algorithms. Objective measures and informal listening tests demonstrated significant improvements over other methods when tested with TIMIT sentences corrupted by various types of colored noise.

A generalized subspace approach is proposed for enhancement of speech corrupted by colored noise. The proposed approach is based on the simultaneous diagonalization of the clean speech and noise covariance matrices, which is shown to be a generalization of the approach proposed by Ephraim and Van Trees for white noise. Objective and subjective measures demonstrated significant improvements over other subspace-based methods when tested with sentences corrupted with speech-shaped noise and multi-talker babble.

The spectral subtraction method is a well-known noise reduction technique. Most implementations and variations of the basic technique advocate subtraction of the noise spectrum estimate over the entire speech spectrum. However, real world noise is mostly colored and does not affect the speech signal uniformly over the entire spectrum. In this paper, we propose a multi-band spectral subtraction approach which takes into account the fact that colored noise affects the speech spectrum differently at various frequencies. This method outperforms the standard power spectral subtraction method resulting in superior speech quality and largely reduced musical noise.

A perceptually motivated subspace based approach is proposed for enhancement of speech corrupted by colored noise.

A new subspace approach is proposed for enhancement of speech corrupted by colored noise. The proposed approach is based on the simultaneous diagonalization of the clean speech and noise covariance matrices, which leads to an optimal linear estimator that minimizes speech distortion subject to the noise distortion being below a set threshold. The proposed approach is shown to be a generalization of the approach proposed by Ephraim and Van Trees (1995) for white noise. Objective and subjective measures demonstrated significant improvements over other subspace-based methods when tested with sentences corrupted with speech-shaped noise and multitalker babble.

Most speech enhancement techniques do not perform very well in correlated or colored noise, as they assume that noise and speech are not correlated. In this paper, we propose a method, based on spectral subtraction, which takes into account possible correlation between noise and speech. Objective measures showed that the proposed method outperformed the power spectral subtraction method resulting in better speech quality and reduced levels of musical noise. Further enhancements in speech quality were obtained by applying a perceptual weighting function (estimated using a psychoacoustics model) that was designed to minimize noise distortion.

A laboratory speech processor based on Motorola's DSP56002 chip was developed for cochlear implant research. The hardware of the speech processor is described in detail and the safety issues associated with electrical stimulation of the auditory nerve are discussed. The speech processor is capable of providing high-rate stimulation to six electrodes using short biphasic pulses presented either simultaneously or in an interleaved fashion. Different speech processing algorithms including the continuous interleaved sampling (CIS) strategy were implemented in this processor and tested successfully with cochlear implant patients.

A speech recognition system based on an Internet client-server model is presented in this paper. A Java applet records the voice at the client computer, sends the recorded speech file over the Internet, and the server computer recognizes the speech and displays the recognized text back to the user. Using this structure, an isolated digit recognition application was realized.

Two class separability criteria based on the divergence measure are proposed to improve speech recognition performance. The average and weighted average divergence measures are used as criteria for finding a transformation matrix which maps the original features into a more discriminative subspace. Results are presented for a highly confusable task.

Alphabet recognition is needed in many applications for retrieving information associated with the spelling of a name, such as telephone numbers, addresses, etc. This is a difficult recognition task due to the acoustic similarities existing between letters in the alphabet (e.g., the E-set letters). This paper presents the development of a high-performance alphabet recognizer that has been evaluated on studio quality as well as on telephone-bandwidth speech. Unlike previously proposed systems, the alphabet recognizer presented is based on context-dependent phoneme hidden Markov models (HMMs), which have been found to outperform whole-word models by as much as 8%. The proposed recognizer incorporates a series of new approaches to tackle the problems associated with the confusions occurring between the stop consonants in the E-set and the confusions between the nasals (i.e., letters M and N). First, a new feature representation is proposed for improved stop consonant discrimination, and second, two subspace approaches are proposed for improved nasal discrimination. The subspace approach was found to yield a 45% error-rate reduction in nasal discrimination. Various other techniques are also proposed, yielding a 97.3% speaker-independent performance on alphabet recognition and 95% speaker-independent performance on E-set recognition, A telephone alphabet recognizer was also developed using context-dependent HMMs. When tested on the recognition of 300 last names (which are contained in a list of 50,000 common last names) spelled by 300 speakers, the recognizer achieved 91.7% correct letter recognition with 1.1% letter insertions.

Alphabet recognition is known to be a difficult task due to the acoustic similarities among different letters, especially letters in the E-set. Recognition systems based on whole-word Hidden Markov Models (HMM) perform poorly on this task due to the inability of the models to capture fine phonetic details, especially details occurring within segments of short duration. Letters B and D, for example, differ mainly in the 10-20 msec segment prior to vowel onset. In this paper, we use context-dependent phoneme-based HMMs to capture the fine phonetic detail that is required to discriminate such a confusable vocabulary. Our results reveal that context-dependent modeling gives about 9% improvement on speaker-independent performance over whole-word modeling, and an 18% improvement on the E-set. Furthermore, using an improved spectral representation of the stop consonants in the E-set, an additional 6% improvement in the E-set can be achieved. Our best speaker-independent E-set performance over 15 speakers is 90.3%, with overall alphabet recognition of 94.1%.>

A block modified covariance algorithm (BMCA) is proposed for autoregressive (AR) parametric spectral estimation. The BMCA can be implemented either in the time or in the frequency domain with the latter being more efficient in high-order AR estimation. Although the BMCA operates on a block of data it allows for both block and sequential (sample-by-sample) updates. The performance of the algorithm is evaluated in terms of computational complexity and convergence characteristics. Results are given for narrowband and broadband processes.>

A transform-based system is described for vector quantizing speech at and below 1200 b/s. A fixed-dimension codebook is used to quantize low-resolution harmonic components of speech. To reduce the errors of the low-resolution process, transform components, within the baseband and adjacent to the pitch-harmonic components, are also encoded. Speech produced at 1200 b/s using this method was found to be intelligible, particularly for female speakers. This is because the number of harmonic components required for female (high-pitch) speakers is generally smaller than the number required for male (low-pitch) speakers.>