Speech Dereverberation Based on Maximum-Likelihood Estimation With Time-Varying Gaussian Source Model

Distant acquisition of acoustic signals in an enclosed space often produces reverberant components due to acoustic reflections in the room. Speech dereverberation is in general desirable when the signal is acquired through distant microphones in such applications as hands-free speech recognition, teleconferencing, and meeting recording. This paper proposes a new speech dereverberation approach based on a statistical speech model. A time-varying Gaussian source model (TVGSM) is introduced as a model that represents the dynamic short time characteristics of nonreverberant speech segments, including the time and frequency structures of the speech spectrum. With this model, dereverberation of the speech signal is formulated as a maximum-likelihood (ML) problem based on multichannel linear prediction, in which the speech signal is recovered by transforming the observed signal into one that is probabilistically more like nonreverberant speech. We first present a general ML solution based on TVGSM, and derive several dereverberation algorithms based on various source models. Specifically, we present a source model consisting of a finite number of states, each of which is manifested by a short time speech spectrum, defined by a corresponding autocorrelation (AC) vector. The dereverberation algorithm based on this model involves a finite collection of spectral patterns that form a codebook. We confirm experimentally that both the time and frequency characteristics represented in the source models are very important for speech dereverberation, and that the prior knowledge represented by the codebook allows us to further improve the dereverberated speech quality. We also confirm that the quality of reverberant speech signals can be greatly improved in terms of the spectral shape and energy time-pattern distortions from simply a short speech signal using a speaker-independent codebook.      

Suppression of Late Reverberation Effect on Speech Signal Using Long-Term Multiple-step Linear Prediction

A speech signal captured by a distant microphone is generally smeared by reverberation, which severely degrades automatic speech recognition (ASR) performance. One way to solve this problem is to dereverberate the observed signal prior to ASR. In this paper, a room impulse response is assumed to consist of three parts: a direct-path response, early reflections and late reverberations. Since late reverberations are known to be a major cause of ASR performance degradation, this paper focuses on dealing with the effect of late reverberations. The proposed method first estimates the late reverberations using long-term multi-step linear prediction, and then reduces the late reverberation effect by employing spectral subtraction. The algorithm provided good dereverberation with training data corresponding to the duration of one speech utterance, in our case, less than 6 s. This paper describes the proposed framework for both single-channel and multichannel scenarios. Experimental results showed substantial improvements in ASR performance with real recordings under severe reverberant conditions.      

Harmonicity-Based Blind Dereverberation for Single-Channel Speech Signals

The distant acquisition of acoustic signals in an enclosed space often produces reverberant artifacts due to the room impulse response. Speech dereverberation is desirable in situations where the distant acquisition of acoustic signals is involved. These situations include hands-free speech recognition, teleconferencing, and meeting recording, to name a few. This paper proposes a processing method, named Harmonicity-based dEReverBeration (HERB), to reduce the amount of reverberation in the signal picked up by a single microphone. The method makes extensive use of harmonicity, a unique characteristic of speech, in the design of a dereverberation filter. In particular, harmonicity enhancement is proposed and demonstrated as an effective way of estimating a filter that approximates an inverse filter corresponding to the room impulse response. Two specific harmonicity enhancement techniques are presented and compared; one based on an average transfer function and the other on the minimization of a mean squared error function. Prototype HERB systems are implemented by introducing several techniques to improve the accuracy of dereverberation filter estimation, including time warping analysis. Experimental results show that the proposed methods can achieve high-quality speech dereverberation, when the reverberation time is between 0.1 and 1.0 s, in terms of reverberation energy decay curves and automatic speech recognition accuracy     

Efficient blind dereverberation and echo cancellation based on independent component analysis for actual acoustic signals

This letter presents a new algorithm for blind dereverberation and echo cancellation based on independent component analysis (ICA) for actual acoustic signals. We focus on frequency domain ICA (FD-ICA) because its computational cost and speed of learning convergence are sufficiently reasonable for practical applications such as hands-free speech recognition. In applying conventional FD-ICA as a preprocessing of automatic speech recognition in noisy environments, one of the most critical problems is how to cope with reverberations. To extract a clean signal from the reverberant observation, we model the separation process in the short-time Fourier transform domain and apply the multiple input/output inverse-filtering theorem (MINT) to the FD-ICA separation model. A naive implementation of this method is computationally expensive, because its time complexity is the second order of reverberation time. Therefore, the main issue in dereverberation is to reduce the high computational cost of ICA. In this letter, we reduce the computational complexity to the linear order of the reverberation time by using two techniques: (1) a separation model based on the independence of delayed observed signals with MINT and (2) spatial sphering for preprocessing. Experiments show that the computational cost grows in proportion to the linear order of the reverberation time and that our method improves the word correctness of automatic speech recognition by 10 to 20 points in a RT??= 670 ms reverberant environment.     

Combination of bottleneck feature extraction and dereverberation for distant-talking speech recognition

The performance of speech recognition in distant-talking environments is severely degraded by the reverberation that can occur in enclosed spaces (e.g., meeting rooms). To mitigate this degradation, dereverberation techniques such as network structure-based denoising autoencoders and multi-step linear prediction are used to improve the recognition accuracy of reverberant speech. Regardless of the reverberant conditions, a novel discriminative bottleneck feature extraction approach has been demonstrated to be effective for speech recognition under a range of conditions. As bottleneck feature extraction is not primarily designed for dereverberation, we are interested in whether it can compensate for other carefully designed dereverberation approaches. In this paper, we propose three schemes covering both front-end processing (cascaded combination and parallel combination) and back-end processing (system combination). Each of these schemes integrates bottleneck feature extraction with dereverberation. The effectiveness of these schemes is evaluated via a series of experiments using the REVERB challenge dataset.     

Static and Dynamic Variance Compensation for Recognition of Reverberant Speech With Dereverberation Preprocessing

The performance of automatic speech recognition is severely degraded in the presence of noise or reverberation. Much research has been undertaken on noise robustness. In contrast, the problem of the recognition of reverberant speech has received far less attention and remains very challenging. In this paper, we use a dereverberation method to reduce reverberation prior to recognition. Such a preprocessor may remove most reverberation effects. However, it often introduces distortion, causing a dynamic mismatch between speech features and the acoustic model used for recognition. Model adaptation could be used to reduce this mismatch. However, conventional model adaptation techniques assume a static mismatch and may therefore not cope well with a dynamic mismatch arising from dereverberation. This paper proposes a novel adaptation scheme that is capable of managing both static and dynamic mismatches. We introduce a parametric model for variance adaptation that includes static and dynamic components in order to realize an appropriate interconnection between dereverberation and a speech recognizer. The model parameters are optimized using adaptive training implemented with the expectation maximization algorithm. An experiment using the proposed method with reverberant speech for a reverberation time of 0.5 s revealed that it was possible to achieve an 80% reduction in the relative error rate compared with the recognition of dereverberated speech (word error rate of 31%), and the final error rate was 5.4%, which was obtained by combining the proposed variance compensation and MLLR adaptation.     

Integrated Speech Enhancement Method Using Noise Suppression and Dereverberation

This paper proposes a method for enhancing speech signals contaminated by room reverberation and additive stationary noise. The following conditions are assumed. 1) Short-time spectral components of speech and noise are statistically independent Gaussian random variables. 2) A room's convolutive system is modeled as an autoregressive system in each frequency band. 3) A short-time power spectral density of speech is modeled as an all-pole spectrum, while that of noise is assumed to be time-invariant and known in advance. Under these conditions, the proposed method estimates the parameters of the convolutive system and those of the all-pole speech model based on the maximum likelihood estimation method. The estimated parameters are then used to calculate the minimum mean square error estimates of the speech spectral components. The proposed method has two significant features. 1) The parameter estimation part performs noise suppression and dereverberation alternately. (2) Noise-free reverberant speech spectrum estimates, which are transferred by the noise suppression process to the dereverberation process, are represented in the form of a probability distribution. This paper reports the experimental results of 1500 trials conducted using 500 different utterances. The reverberation time RT₆₀ was 0.6 s, and the reverberant signal to noise ratio was 20, 15, or 10 dB. The experimental results show the superiority of the proposed method over the sequential performance of the noise suppression and dereverberation processes.     

Joint Dereverberation and Residual Echo Suppression of Speech Signals in Noisy Environments

Hands-free devices are often used in a noisy and reverberant environment. Therefore, the received microphone signal does not only contain the desired near-end speech signal but also interferences such as room reverberation that is caused by the near-end source, background noise and a far-end echo signal that results from the acoustic coupling between the loudspeaker and the microphone. These interferences degrade the fidelity and intelligibility of near-end speech. In the last two decades, postfilters have been developed that can be used in conjunction with a single microphone acoustic echo canceller to enhance the near-end speech. In previous works, spectral enhancement techniques have been used to suppress residual echo and background noise for single microphone acoustic echo cancellers. However, dereverberation of the near-end speech was not addressed in this context. Recently, practically feasible spectral enhancement techniques to suppress reverberation have emerged. In this paper, we derive a novel spectral variance estimator for the late reverberation of the near-end speech. Residual echo will be present at the output of the acoustic echo canceller when the acoustic echo path cannot be completely modeled by the adaptive filter. A spectral variance estimator for the so-called late residual echo that results from the deficient length of the adaptive filter is derived. Both estimators are based on a statistical reverberation model. The model parameters depend on the reverberation time of the room, which can be obtained using the estimated acoustic echo path. A novel postfilter is developed which suppresses late reverberation of the near-end speech, residual echo and background noise, and maintains a constant residual background noise level. Experimental results demonstrate the beneficial use of the developed system for reducing reverberation, residual echo, and background noise.      

Constructing accurate and robust HMM/GMM models for an Arabic speech recognition system

Conventional Hidden Markov Model (HMM) based Automatic Speech Recognition (ASR) systems generally utilize cepstral features as acoustic observation and phonemes as basic linguistic units. Some of the most powerful features currently used in ASR systems are Mel-Frequency Cepstral Coefficients (MFCCs). Speech recognition is inherently complicated due to the variability in the speech signal which includes within- and across-speaker variability. This leads to several kinds of mismatch between acoustic features and acoustic models and hence degrades the system performance. The sensitivity of MFCCs to speech signal variability motivates many researchers to investigate the use of a new set of speech feature parameters in order to make the acoustic models more robust to this variability and thus improve the system performance. The combination of diverse acoustic feature sets has great potential to enhance the performance of ASR systems. This paper is a part of ongoing research efforts aspiring to build an accurate Arabic ASR system for teaching and learning purposes. It addresses the integration of complementary features into standard HMMs for the purpose to make them more robust and thus improve their recognition accuracies. The complementary features which have been investigated in this work are voiced formants and Pitch in combination with conventional MFCC features. A series of experimentations under various combination strategies were performed to determine which of these integrated features can significantly improve systems performance. The Cambridge HTK tools were used as a development environment of the system and experimental results showed that the error rate was successfully decreased, the achieved results seem very promising, even without using language models.     

New Insights Into the MVDR Beamformer in Room Acoustics

The minimum variance distortionless response (MVDR) beamformer, also known as Capon's beamformer, is widely studied in the area of speech enhancement. The MVDR beamformer can be used for both speech dereverberation and noise reduction. This paper provides new insights into the MVDR beamformer. Specifically, the local and global behavior of the MVDR beamformer is analyzed and novel forms of the MVDR filter are derived and discussed. In earlier works it was observed that there is a tradeoff between the amount of speech dereverberation and noise reduction when the MVDR beamformer is used. Here, the tradeoff between speech dereverberation and noise reduction is analyzed thoroughly. The local and global behavior, as well as the tradeoff, is analyzed for different noise fields such as, for example, a mixture of coherent and non-coherent noise fields, entirely non-coherent noise fields and diffuse noise fields. It is shown that maximum noise reduction is achieved when the MVDR beamformer is used for noise reduction only. The amount of noise reduction that is sacrificed when complete dereverberation is required depends on the direct-to-reverberation ratio of the acoustic impulse response between the source and the reference microphone. The performance evaluation supports the theoretical analysis and demonstrates the tradeoff between speech dereverberation and noise reduction. When desiring both speech dereverberation and noise reduction, the results also demonstrate that the amount of noise reduction that is sacrificed decreases when the number of microphones increases.      

Artificial neural networks as speech recognisers for dysarthric speech: Identifying the best-performing set of MFCC parameters and studying a speaker-independent approach

Dysarthria is a neurological impairment of controlling the motor speech articulators that compromises the speech signal. Automatic Speech Recognition (ASR) can be very helpful for speakers with dysarthria because the disabled persons are often physically incapacitated. Mel-Frequency Cepstral Coefficients (MFCCs) have been proven to be an appropriate representation of dysarthric speech, but the question of which MFCC-based feature set represents dysarthric acoustic features most effectively has not been answered. Moreover, most of the current dysarthric speech recognisers are either speaker-dependent (SD) or speaker-adaptive (SA), and they perform poorly in terms of generalisability as a speaker-independent (SI) model. First, by comparing the results of 28 dysarthric SD speech recognisers, this study identifies the best-performing set of MFCC parameters, which can represent dysarthric acoustic features to be used in Artificial Neural Network (ANN)-based ASR. Next, this paper studies the application of ANNs as a fixed-length isolated-word SI ASR for individuals who suffer from dysarthria. The results show that the speech recognisers trained by the conventional 12 coefficients MFCC features without the use of delta and acceleration features provided the best accuracy, and the proposed SI ASR recognised the speech of the unforeseen dysarthric evaluation subjects with word recognition rate of 68.38%.     

Robust Speech Dereverberation Using Multichannel Blind Deconvolution With Spectral Subtraction

A robust dereverberation method is presented for speech enhancement in a situation requiring adaptation where a speaker shifts his/her head under reverberant conditions causing the impulse responses to change frequently. We combine correlation-based blind deconvolution with modified spectral subtraction to improve the quality of inverse-filtered speech degraded by the estimation error of inverse filters obtained in practice. Our method computes inverse filters by using the correlation matrix between input signals that can be observed without measuring room impulse responses. Inverse filtering reduces early reflection, which has most of the power of the reverberation, and then, spectral subtraction suppresses the tail of the inverse-filtered reverberation. The performance of our method in adaptation is demonstrated by experiments using measured room impulse responses. The subjective results indicated that this method provides superior speech quality to each of the individual methods: blind deconvolution and spectral subtraction.     

Performance analysis of ASR system in hybrid DNN-HMM framework using a PWL euclidean activation function

Automatic Speech Recognition (ASR) is the process of mapping an acoustic speech signal into a human readable text format. Traditional systems exploit the Acoustic Component of ASR using the Gaussian Mixture Model- Hidden Markov Model (GMM-HMM) approach.Deep NeuralNetwork (DNN) opens up new possibilities to overcome the shortcomings of conventional statistical algorithms. Recent studies modeled the acoustic component of ASR system using DNN in the so called hybrid DNN-HMM approach. In the context of activation functions used to model the non-linearity in DNN, Rectified Linear Units (ReLU) and maxout units are mostly used in ASR systems. This paper concentrates on the acoustic component of a hybrid DNN-HMM system by proposing an efficient activation function for the DNN network. Inspired by previous works, euclidean norm activation function is proposed to model the non-linearity of the DNN network. Such non-linearity is shown to belong to the family of Piecewise Linear (PWL) functions having distinct features. These functions can capture deep hierarchical features of the pattern. The relevance of the proposal is examined in depth both theoretically and experimentally. The performance of the developed ASR system is evaluated in terms of Phone Error Rate (PER) using TIMIT database. Experimental results achieve a relative increase in performance by using the proposed function over conventional activation functions.     

Precise Dereverberation Using Multichannel Linear Prediction

In this paper, we discuss the numerical problems posed by the previously reported LInear-predictive Multi-input Equalization (LIME) algorithm when dealing with dereverberation of long room transfer functions (RTF). The LIME algorithm consists of two steps. First, a speech residual is calculated using multichannel linear prediction. The residual is free from the room reverberation effect but it is also excessively whitened because the average speech characteristics have been removed. In the second step, LIME estimates such average speech characteristics to compensate for the excessive whitening. When multiple microphones are used, the speech characteristics are common to all microphones whereas the room reverberation differs for each microphone. LIME estimates the average speech characteristics as the characteristics that are common to all the microphones. Therefore, LIME relies on the hypothesis that there are no zeros common to all channels. However, it is known that RTFs have a large number of zeros close to the unit circle on the z-plane. Consequently, the zeros of the RTFs are distributed in the same regions of the z-plane and, if an insufficient number of microphones are used, the channels would present numerically overlapping zeros. In such a case, the dereverberation algorithm would perform poorly. We discuss the influence of overlapping zeros on the dereverberation performance of LIME. Spatial information can be used to deal with the problem of overlapping zeros. By increasing the number of microphones, the number of overlapping zeros decreases and the dereverberation performance is improved. We also examine the use of cepstral mean normalization for post-processing to reduce the remaining distortions caused by the overlapping zeros     

Speaker-independent ASR for Modern Standard Arabic: effect of regional accents

This paper deals with speaker-independent Automatic Speech Recognition (ASR) system for continuous speech. This ASR system has been developed for Modern Standard Arabic (MSA) using recordings of six regions taken from ALGerian Arabic Speech Database (ALGASD), and has been designed by using Hidden Markov Models. The main purpose of this study is to investigate the effect of regional accent on speech recognition rates. First, the experiment assessed the general performance of the model for the data speech of six regions, details of the recognition results are performed to observe the deterioration of the performance of the ASR according to the regional variation included in the speech material. The results have shown that the ASR performance is clearly impacted by the regional accents of the speakers.     

Unsupervised Adaptation of Categorical Prosody Models for Prosody Labeling and Speech Recognition

Automatic speech recognition (ASR) systems rely almost exclusively on short-term segment-level features (MFCCs), while ignoring higher level suprasegmental cues that are characteristic of human speech. However, recent experiments have shown that categorical representations of prosody, such as those based on the Tones and Break Indices (ToBI) annotation standard, can be used to enhance speech recognizers. However, categorical prosody models are severely limited in scope and coverage due to the lack of large corpora annotated with the relevant prosodic symbols (such as pitch accent, word prominence, and boundary tone labels). In this paper, we first present an architecture for augmenting a standard ASR with symbolic prosody. We then discuss two novel, unsupervised adaptation techniques for improving, respectively, the quality of the linguistic and acoustic components of our categorical prosody models. Finally, we implement the augmented ASR by enriching ASR lattices with the adapted categorical prosody models. Our experiments show that the proposed unsupervised adaptation techniques significantly improve the quality of the prosody models; the adapted prosodic language and acoustic models reduce binary pitch accent (presence versus absence) classification error rate by 13.8% and 4.3%, respectively (relative to the seed models) on the Boston University Radio News Corpus, while the prosody-enriched ASR exhibits a 3.1% relative reduction in word error rate (WER) over the baseline system.     

Opening the Knowledge Dam: Speech Recognition for Video Search

Automatic Speech Recognition (ASR) may increase access to spoken information captured in videos. ASR is needed, especially for online academic video lectures that gradually replace class lectures and traditional textbooks. This conceptual article examines how technological barriers to ASR in under-resourced languages impair accessibility to video content and demonstrates it with the empirical findings of Hebrew ASR evaluations. We compare ASR with Optical Character Recognition (OCR) as facilitating access to textual and speech content and show their current performance in under-resourced languages. We target ASR of under-resourced languages as the main barrier to searching academic video lectures. We further show that information retrieval technologies, such as smart video players that combine both ASR and OCR capacities, must come to the fore once ASR technologies have matured. Therefore, suggesting that the current state of information retrieval from video lectures in under-resourced languages is equivalent to a knowledge dam.     

Sub-band temporal modulation envelopes and their normalization for automatic speech recognition in reverberant environments

Automatic speech recognition (ASR) in reverberant environments is still a challenging task. In this study, we propose a robust feature-extraction method on the basis of the normalization of the sub-band temporal modulation envelopes (TMEs). The sub-band TMEs were extracted using a series of constant bandwidth band-pass filters with Hilbert transforms followed by low-pass filtering. Based on these TMEs, the modulation spectrums in both clean and reverberation spaces are transformed to a reference space by using modulation transfer functions (MTFs), wherein the MTFs are estimated as the measure of the modulation transfer effect on the sub-band TMEs between the clean, reverberation, and reference spaces. By using the MTFs on the modulation spectrum, it is supposed that the difference on the modulation spectrum caused by the difference of the recording environments is removed. Based on the normalized modulation spectrum, inverse Fourier transform was conducted to restore the sub-band TMEs by retaining their original phase information. We tested the proposed method on speech recognition experiments in a reverberant room with differing speaker to microphone distance (SMD). For comparison, the recognition performance of using the traditional Mel frequency cepstral coefficients with mean and variance normalization was used as the baseline. The experimental results showed that by averaging the results for SMDs from 50 cm to 400 cm, we obtained a 44.96% relative improvement by only using sub-band TME processing, and obtained a further 15.68% relative improvement by performing the normalization on the modulation spectrum of the sub-band TMEs. In all, we obtained a 53.59% relative improvement, which was better than using other temporal filtering and normalization methods.     

A stereophonic acoustic signal extraction scheme for noisy and reverberant environments

In this contribution, a novel two-channel acoustic front-end for robust automatic speech recognition in adverse acoustic environments with nonstationary interference and reverberation is proposed. From a MISO system perspective, a statistically optimum source signal extraction scheme based on the multichannel Wiener filter (MWF) is discussed for application in noisy and underdetermined scenarios. For free-field and diffuse noise conditions, this optimum scheme reduces to a Delay & Sum beamformer followed by a single-channel Wiener postfilter. Scenarios with multiple simultaneously interfering sources and background noise are usually modeled by a diffuse noise field. However, in reality, the free-field assumption is very weak because of the reverberant nature of acoustic environments. Therefore, we propose to estimate this simplified MWF solution in each frequency bin separately to cope with reverberation. We show that this approach can very efficiently be realized by the combination of a blocking matrix based on semi-blind source separation (‘directional BSS’), which provides a continuously updated reference of all undesired noise and interference components separated from the desired source and its reflections, and a single-channel Wiener postfilter. Moreover, it is shown, how the obtained reference signal of all undesired components can efficiently be used to realize the Wiener postfilter, and at the same time, generalizes well-known postfilter realizations. The proposed front-end and its integration into an automatic speech recognition (ASR) system are analyzed and evaluated in noisy living-room-like environments according to the PASCAL CHiME challenge. A comparison to a simplified front-end based on a free-field assumption shows that the introduced system substantially improves the speech quality and the recognition performance under the considered adverse conditions.     

Modulation Spectral Features for Robust Far-Field Speaker Identification

In this paper, auditory inspired modulation spectral features are used to improve automatic speaker identification (ASI) performance in the presence of room reverberation. The modulation spectral signal representation is obtained by first filtering the speech signal with a 23-channel gammatone filterbank. An eight-channel modulation filterbank is then applied to the temporal envelope of each gammatone filter output. Features are extracted from modulation frequency bands ranging from 3-15 H z and are shown to be robust to mismatch between training and testing conditions and to increasing reverberation levels. To demonstrate the gains obtained with the proposed features, experiments are performed with clean speech, artificially generated reverberant speech, and reverberant speech recorded in a meeting room. Simulation results show that a Gaussian mixture model based ASI system, trained on the proposed features, consistently outperforms a baseline system trained on mel-frequency cepstral coefficients. For multimicrophone ASI applications, three multichannel score combination and adaptive channel selection techniques are investigated and shown to further improve ASI performance.