For a better understanding of the mechanisms underlying speech perception and the contribution of different signal features, computational models of speech recognition have a long tradition in hearing research.Due to the diverse range of situations in which speech needs to be recognized, these models need to be generalizable across many acoustic conditions, speakers, and languages.This contribution examines the importance of different features for speech recognition predictions of plain and Lombard speech for English in comparison to Cantonese in stationary and modulated noise.While Cantonese is a tonal language that encodes information in spectro-temporal features, the Lombard effect is known to be associated with spectral changes in the speech signal.These contrasting properties of tonal languages and the Lombard effect form an interesting basis for the assessment of speech recognition models.Here, an automatic speech recognition-based (ASR) model using spectral or spectro-temporal features is evaluated with empirical data.The results indicate that spectro-temporal features are crucial in order to predict the speaker-specific speech recognition threshold SRT50 in both Cantonese and English as well as to account for the improvement of speech recognition in modulated noise, while effects due to Lombard speech can already be predicted by spectral features.

The reverberation time (RT) and the early-to-late reverberation ratio (ELR) are two key parameters commonly used to characterize acoustic room environments. In contrast to conventional blind estimation methods that process the two parameters separately, we propose a model for joint estimation to predict the RT and the ELR simultaneously from single-channel speech signals from either full-band or sub-band frequency data, which is referred to as joint room parameter estimator (jROPE). An artificial neural network is employed to learn the mapping from acoustic observations to the RT and the ELR classes. Auditory-inspired acoustic features obtained by temporal modulation filtering of the speech time-frequency representations are used as input for the neural network. Based on an in-depth analysis of the dependency between the RT and the ELR, a two-dimensional (RT, ELR) distribution with constrained boundaries is derived, which is then exploited to evaluate four different configurations for jROPE. Experimental results show that-in comparison to the single-task ROPE system which individually estimates the RT or the ELR-jROPE provides improved results for both tasks in various reverberant and (diffuse) noisy environments. Among the four proposed joint types, the one incorporating multi-task learning with shared input and hidden layers yields the best estimation accuracies on average. When encountering extreme reverberant conditions with RTs and ELRs lying beyond the derived (RT, ELR) distribution, the type considering RT and ELR as a joint parameter performs robustly, in particular. From state-of-the-art algorithms that were tested in the acoustic characterization of environments challenge, jROPE achieves comparable results among the best for all individual tasks (RT and ELR estimation from full-band and sub-band signals).

S.981-985

Room acoustic parameters that characterize acoustic environments can help to improve signal enhancement algorithms such as for dereverberation, or automatic speech recognition by adapting models to the current parameter set. The reverberation time (RT) and the early-to-late reverberation ratio (ELR) are two key parameters. In this paper, we propose a blind ROom Parameter Estimator (ROPE) based on an artificial neural network that learns the mapping to discrete ranges of the RT and the ELR from single-microphone speech signals. Auditory-inspired acoustic features are used as neural network input, which are generated by a temporal modulation filter bank applied to the speech time-frequency representation. ROPE performance is analyzed in various reverberant environments in both clean and noisy conditions for both fullband and subband RT and ELR estimations. The importance of specific temporal modulation frequencies is analyzed by evaluating the contribution of individual filters to the ROPE performance. Experimental results show that ROPE is robust against different variations caused by room impulse responses (measured versus simulated), mismatched noise levels, and speech variability reflected through different corpora. Compared to state-of-the-art algorithms that were tested in the acoustic characterisation of environments (ACE) challenge, the ROPE model is the only one that is among the best for all individual tasks (RT and ELR estimation from fullband and subband signals). Improved fullband estimations are even obtained by ROPE when integrating speech-related frequency subbands. Furthermore, the model requires the least computational resources with a real time factor that is at least two times faster than competing algorithms. Results are achieved with an average observation window of 3 s, which is important for real-time applications.

This paper evaluates neural network (NN) based systems and compares them to Gaussian mixture model (GMM) and hidden Markov model (HMM) approaches for acoustic scene classification (SC) and polyphonic acoustic event detection (AED) that are applied to data of the “Detection and Classification of Acoustic Scenes and Events 2016” (DCASE'16) challenge, task 1 and task 3, respectively. For both tasks, the use of deep neural networks (DNNs) and features based on an amplitude modulation filterbank and a Gabor filterbank (GFB) are evaluated and compared to standard approaches. For SC, additionally a time-delay NN approach is proposed that enables analysis of long contextual information similar to recurrent NNs but with training efforts comparable to conventional DNNs. The SC system proposed for task 1 of the DCASE'16 challenge attains a recognition accuracy of 77.5%, which is 5.6% higher compared to the DCASE'16 baseline system. For the AED task, DNNs are adopted in tandem and hybrid approaches, i.e., as part of HMM-based systems. These systems are evaluated for the polyphonic data of task 3 from the DCASE'16 challenge. Several strategies to address the issue of polyphony are considered. It is shown that DNN-based systems perform less accurate than the traditional systems for this task. Best results are achieved using GFB features in combination with a multiclass GMM-HMM back end.

The segregation of concurrent speakers and other sound sources is an important ability of the human auditory system, but is missing in most current systems for automatic speech recognition (ASR), resulting in a large gap between human and machine performance. This study combines processing related to peripheral and cortical stages of the auditory pathway: A physiologically motivated binaural model estimates the positions of moving speakers to enhance the desired speech signal. Second, signals are converted to spectro-temporal Gabor features that resemble cortical speech representations and which have been shown to improve ASR in noisy conditions. Spectro-temporal Gabor features improve recognition results in all acoustic conditions under consideration compared with Mel-frequency cepstral coefficients. Binaural processing results in lower word error rates (WERs) in acoustic scenes with a concurrent speaker, whereas monaural processing should be preferred in the presence of a stationary masking noise. In-depth analysis of binaural processing identifies crucial processing steps such as localization of sound sources and estimation of the beamformer's noise coherence matrix, and shows how much each processing step affects the recognition performance in acoustic conditions with different complexity.

Sensorineural hearing loss typically results in elevated thresholds and steepened loudness growth significantly conditioned by a damage of outer hair cells (OHC). In hearing aids, amplification and dynamic compression aim at widening the limited available dynamic range. However, speech perception particularly in complex acoustic scenes often remains difficult. Here, a physiologically motivated, fast acting, model-based dynamic compression algorithm (MDC) is introduced which aims at restoring the behaviorally estimated basilar membrane input-output (BM I/O) function in normal-hearing listeners. A system-specific gain prescription rule is suggested, based on the same model BM I/O function and a behavioral estimate of the individual OHC loss. Cochlear off-frequency component suppression is mimicked using an instantaneous frequency (IF) estimate. Increased loudness as a consequence of widened filters in the impaired system is considered in a further compensation stage. In an extended version, a subband estimate of the speech presence probability (MDC+SPP) additionally provides speech-selective amplification in stationary noise. Instrumental evaluation revealed that the IF control enhances the spectral contrast of vowels and benefits in quality predictions at higher signal-to-noise ratios (SNRs) were observed. Compared with a conventional multiband dynamic compressor, MDC achieved objective quality and intelligibility benefits for a competing talker at lower SNRs. MDC+SPP outperformed the conventional compressor in the quality predictions and reached comparable instrumental speech intelligibility as achieved with linear amplification. The proposed algorithm provides a first promising basis for auditory model-based compression with signal-type- and bandwidth-dependent gains.

Inspired by physiological studies on the human auditory system and by results from psychoacoustics, an amplitude modulation filter bank (AMFB) has been developed and successfully applied to feature extraction for automatic speech recognition (ASR) in earlier work. Here, we address the question as to which amplitude modulation (AM) frequency decomposition leads to optimal ASR performance by proposing a parameterized functional relationship between modulation center frequency and modulation bandwidth. Word error rates (WERs) of ASR experiments with 1551 different AMFBs are systematically evaluated and compared, resulting in the identification of a comparatively narrow range of optimal modulation frequency to modulation bandwidth characteristics. To integrate modulation processing with deep neural network (DNN) acoustic modeling, we propose merging of modulation filter coefficients with DNN weights prior to a final training step and an improved mean-variance normalization scheme for AMFBs. These modifications are shown to result in further reduction of WERs and are indicative of the proposed system's improved generalization ability, when compared across corpora of 100-960 h of data with mismatched training and test conditions. Analysis of DNN-learned temporal AM filtering properties is carried out and implications for the relevance of different modulation regions as well as the relation to psychoacoustic findings are discussed. ASR experiments with the proposed system demonstrate a high degree of robustness against extrinsic acoustic distortions, resulting in, e.g., an average WER of 9.79% on the Aurora-4 task.

The paper describes an automatic speech recognition (ASR) system for the 3rd CHiME challenge that addresses noisy acoustic scenes within public environments. The proposed system includes a multi-channel speech enhancement front-end including a microphone channel failure detection method that is based on cross-comparing the modulation spectra of speech to detect erroneous microphone recordings. The main focus of the submission is the investigation of the amplitude modulation filter bank (AMFB) as a method to extract long-term acoustic cues prior to a Gaussian mixture model (GMM) or deep neural network (DNN) based ASR classifier. It is shown that AMFB features outperform the commonly used frame splicing technique of filter bank features even on a performance optimized ASR challenge system. I.e., temporal analysis of speech by hand-crafted and auditory motivated AMFBs is shown to be more robust compared to a data-driven method based on extracting temporal dynamics with a DNN. Our final ASR system, which additionally includes adaptation of acoustic features to speaker characteristics, achieves an absolute word error rate reduction of approx. 21.53 % relative to the best CHiME-3 baseline system on the "real" test condition.

The human ability to classify acoustic sounds is still unmatched compared to recent methods in machine learning. Psychoacoustic and physiological studies indicate that the auditory system of mammals decomposes audio signals into their acoustic and modulation frequency components prior to further analysis. Since it is known that most linguistic information is coded in amplitude fluctuations, mimicking temporal processing strategies of the auditory system in automatic speech recognition (ASR) promises to increase recognition accuracies. We present an amplitude modulation filter bank (AMFB) that is used as a feature extraction scheme in ASR systems. The time-frequency resolution of the employed FIR filters, i.e., bandwidth and modulation frequency settings, are adopted from a psychophysically inspired model of Dau (1997) that was originally proposed to describe data from human psychoacoustics. Investigations on modulation phase indicate the need for preserving such information in amplitude modulation features. We show that the filter symmetry has an important impact on ASR performance. The proposed feature extraction scheme exhibits significant word error rate (WER) reductions using the Aurora-2, Aurora-4, and REVERB ASR tasks compared to other recent feature extraction methods, such as MFCC, FDLP, and PNCC features. Thereby, AMFB features reveal high robustness against additive noise, different transmission channel characteristics, and room reverberation. Using the Aurora-4 benchmark, for instance, an average WER of 12.33% with raw and 11.31% with bottleneck transformed features is attained, which constitutes a relative improvement of 19.6% and 29.2% over raw MFCC features, respectively.

A perceptual, binaural audio-quality model is introduced. The model was developed for predicting any kinds of perceived spatial quality differences between two audio signals in multi-channel reproduction and audio processing schemes. It employs a recent binaural auditory model as front-end to provide perceptually relevant binaural features for the reference and test audio signal. Correlations between the binaural features of both signals are combined to an overall spatial quality measure by the use of multivariate adaptive regression splines (MARS). Furthermore, a database was generated to train and evaluate the model. The database contains various multi-channel audio signals, which were subjectively assessed in formal listening tests with 15 trained listeners. The results show different model prediction performances depending on the type of quality degradation. Combination of the proposed spatial quality measure with established monaural quality measures improved the predictive power.

Spectro-temporal filtering has been shown to result in features that can help to increase the robustness of automatic speech recognition (ASR) in the past. We replace the spectro-temporal representation used in previous work with spectrograms that incorporate knowledge about the signal processing of the human auditory system and which are derived from Power-Normalized Cepstral Coefficients (PNCCs). 2D-Gabor filters are applied to these spectrograms to extract features evaluated on a noisy digit recognition task. The filter bank is adapted to the new representation by optimizing the spectral modulation frequencies associated with each Gabor function. A comparison of optimized parameters and the spectral modulation of vowels shows a good match between optimized and expected range of frequencies. When processed with a non-linear neural net and combined with PNCCs, Gabor features decrease the error rate compared to the baseline and PNCCs by at least 19%.

Many research efforts in the field of feature extraction for automatic speech recognition are focused on analyzing slow amplitude fluctuations of speech. In this study the importance of spectral and temporal resolution for the amplitude modulation frequency analysis are investigated in order to provide guidance for the appropriate filter design. Therefore, different wavelet and Fourier transform like filter time scales are examined, i.e. The importance of time and frequency separation is compared. The results demonstrate that analyzing three separate amplitude modulation frequency bands of constant absolute bandwidth that cover the range from about 2 to 16 Hz are sufficient for automatic speech recognition.

In this contribution we present a feature extraction method that relies on the modulation-spectral analysis of amplitude fluctuations within sub-bands of the acoustic spectrum by a STFT. The experimental results indicate that the optimal temporal filter extension for amplitude modulation analysis is around 310 ms. It is also demonstrated that the phase information of the modulation spectrum contains important cues for speech recognition. In this context, the advantage of an odd analysis basis function is considered. The best presented features reached a total relative improvement of 53.5% for clean-condition training on Aurora-2. Furthermore, it is shown that modulation features are more robust against room reverberation than conventional cepstral and dynamic features and that they strongly benefit from a high early-to-late energy ratio of the characteristic RIR.

Robust detection of speech embedded in real acoustic background noise is considered using an approach based on subband amplitude modulation spectral (AMS) features and trained discriminative classifiers. Performance is evaluated in particular for situations in which speech is embedded in acoustic backgrounds not presented during classifier training, and for signal-to-noise ratios (SNR) from -10 dB to 20 dB. The results show that (1) Generalization to novel background classes with AMS features yields better performance in 84% of investigated situations, corresponding to an SNR benefit of about 10 dB compared to mel-frequency cepstral coefficient (MFCC) features. (2) On known backgrounds, AMS and MFCCs achieve similar performance, with a small advantage for AMS in negative SNR regimes. (3) Standard voice activity detection (ITU G729.B) performs significantly worse than the classification-based approach.

In an attempt to improve models of human perception, the recognition of phonemes in nonsense utterances was predicted with automatic speech recognition (ASR) in order to analyze its applicability for modeling human speech recognition (HSR) in noise. In the first experiments, several feature types are used as input for an ASR system; the resulting phoneme scores are compared to listening experiments using the same speech data. With conventional training, the highest correlation between predicted and measured recognition was observed for perceptual linear prediction features (r = 0.84). Secondly, a new training paradigm for ASR is proposed with the aim of improving the prediction of phoneme intelligibility. For this perceptual training, the original utterance labels are modified based on the confusions measured in HSR tests. The modified ASR training improved the overall prediction, with the best models (r = 0.89) exceeding those obtained with conventional training

In this study, the effect of speech-intrinsic variabilities such as speaking rate, effort and speaking style on automatic speech recognition (ASR) is investigated. We analyze the influence of such variabilities as well as extrinsic factors (i.e., additive noise) on the most common features in ASR (mel-frequency cepstral coefficients and perceptual linear prediction features) and spectro-temporal Gabor features. MFCCs performed best for clean speech, whereas Gabors were found to be the most robust feature in extrinsic variabilities. Intrinsic variations were found to have a strong impact on error rates. While performance with MFCCs and PLPs was degraded in much the same way, Gabor features exhibit a different sensivity towards these variabilities and are, e.g., well-suited to recognize speech with varying pitch. The results suggest that spectro-temporal and classic features carry complementary information, which could be exploited in feature-stream experiments. Index Terms: automatic speech recognition, speech-intrinsic variabilities, feature extraction, spectro-temporal features

The effect of a binaural coherence-based noise reduction scheme on the feedback stability margin and sound quality in hearing aids has been analyzed. For comparison, a conventional adaptive feedback canceler (AFC) and the combination of the adaptive filter with the binaural coherence filter have been tested. The observed quantities are feedback stability and target signal attenuation. An objective measure of feedback stability, i.e., the added stable gain (ASG) was obtained for a number of algorithmic settings and compared to a subjective measure of feedback stability, the added tolerable gain (ATG). In an attempt to eliminate the subjective bias in estimating the ATG, the ldquounbiased added gainrdquo (UAG) is introduced as a new method. Both, objective and subjective measures give similar results for feedback stability. This allows for a valid comparison across different feedback reduction schemes both in isolation and in combination: whereas the ASG of the coherence filter without combination with AFC is negligible, the results indicate that a robust feedback suppression in hearing aids can be achieved if the benefit of de-correlation and the head-shadow effect in binaural hearing aids is used in an advantageous way. The ASG reaches 23 dB for the best combination at the expense of an average target signal attenuation of 15 dB at critical frequencies. The attribute of the coherence filter is that it adaptively limits the maximum gain before feedback becomes audible. The UAG analysis revealed that subjects used a stable quality criterion across the conditions tested and that the group of subjects covered a large range of individual quality criteria.

In this study a self-steering beamformer with binaural output for a head-worn microphone array is investigated in simulated and real- world conditions. The influence of the underlying sound propagation model on the estimation accuracy of the direction of arrival (DOA) estimation algorithm and the overall performance of the combined DOA-beamformer-system is evaluated. For this, technical performance measures as well as objective quality measures based on perceptual models of the auditory system are used. The self-steering beamformer showed better performance than a beamformer with fixed look-direction for SNR values above -2 dB if the propagation model includes at least a coarse head model.

In order to enhance automatic speech recognition performance in adverse conditions, Gabor features motivated by physiolog-ical measurements in the primary auditory cortex were opti-mized and evaluated. In the Aurora 2 experimental setup such localized, spectro-temporal filters combined with a Tandem sys-tem yield robust performance with a feature set size of 30. Im-proved results can be obtained when using a Hanning window instead of a cut-off Gaussian envelope due to better modulation frequency characteristics. An analysis of complementarity of Gabor and MFCC features shows that errors could be reduced by 55 % with a perfect classifier. In a real world scenario, a rela-tive WER reduction of 15 % compared to a competitive baseline is achieved by combining the feature types, indicating the po-tential of this class of physiologically motivated features. Index Terms: spectro temporal features, automatic speech recognition, Gabor features 1.

In this study speech intelligibility in noise for normal-hearing subjects is predicted by a model that consists of an auditory preprocessing and a speech recognizer. Using a highly systematic speech corpus of phoneme combinations (logatomes) allows the analysis of response rates and confusions of single phonemes. The predicted data is validated by listening tests using the same nonsense speech material. If testing utterances that are not identical to those in training material are used, the psychometric function in noise is predicted with an offset of 13 dB to higher signal-to-noiseratios (SNR). This is consistent with the man-machine performance gap between human speech recognition (HSR) and automatic speech recognition (ASR) [1]. However, this offset reduces to 4 dB in a second model design with identical recordings for training and testing. Furthermore predicted confusion matrices are compared to those of normal-hearing subjects with the second model design.

A comparison between automatic speech recognition (ASR) and human speech recognition (HSR) is performed as prerequisite for identifying sources of errors and improving feature extraction in ASR. HSR and ASR experiments are carried out with the same logatome database which consists of nonsense syllables. Two different kinds of signals are presented to human listeners: First, noisy speech samples are converted to Mel-frequency cepstral coefficients which are resynthesized to speech, with information about voicing and fundamental frequency being discarded. Second, the original signals with added noise are presented, which is used to evaluate the loss of information caused by the process of resynthesis. The analysis also covers the degradation of ASR caused by dialect or accent and shows that different error patterns emerge for ASR and HSR. The information loss induced by the calculation of ASR features has the same effect as a deteriation of the SNR by 10 dB. Index Terms: human speech recognition, automatic speech recognition, dialect, accent, phoneme confusions, MFCC

A new method for the objective assessment and prediction of perceived audio quality is introduced. It represents an expansion of the speech quality measure qC, introduced by Hansen and Kollmeier, and is based on a psychoacoustically validated, quantitative model of the "effective" peripheral auditory processing by Dau et al. To evaluate the audio quality of a given distorted signal relative to a corresponding high-quality reference signal, the auditory model is employed to compute "internal representations" of the signals, which are partly assimilated in order to account for assumed cognitive aspects. The linear cross correlation coefficient of the assimilated internal representations represents the perceptual similarity measure (PSM). PSM shows good correlations with subjective quality ratings if different types of audio signals are considered separately, whereas a better accuracy of signal-independent quality prediction is achieved by a second quality measure PSMtrepresented by the fifth percentile of the sequence of instantaneous audio quality PSM(t). The new measures were evaluated using a large database of subjective listening tests that were originally carried out on behalf of the International Telecommunication Union (ITU) and Moving Pictures Experts Group (MPEG) for the evaluation of various low bit-rate audio codecs. Additional tests with data unknown in the development phase of the model were carried out. Except for linear distortions, the new method shows a higher prediction accuracy than the ITU-R recommendation BS.1387 ("PEAQ") for the tested data

This paper introduces the new OLdenburg LOgatome speech corpus (OLLO) and outlines design considerations during its creation. OLLO is distinct from previous ASR corpora as it specifically targets (1) the fair comparison between human and machine speech recognition performance, and (2) the realistic representation of intrinsic variabilities in speech that are significant for automatic speech recognition (ASR) systems. To enable an unbiased human-machine comparison, OLLO is designed for recognition of individual phonemes that are embedded in logatomes, specifically, three-phoneme sequences with no semantic information. A balanced set of target-phonemes important for human and automatic speech recognition has been chosen, drawing on pilot ASR studies and cross-fertilization from the field of human speech intelligibility testing. Several intrinsic variabilities in speech are represented in OLLO, by recording from 40 speakers from four German dialect regions, and by covering six articulation characteristics. Results from preliminary phonetic time-labeling and ASR experiments are promising and consistent with corpus variabilities. 1.

A single-microphone noise suppression algorithm is described that is based on a novel approach for the estimation of the signal-to-noise ratio (SNR) in different frequency channels: The input signal is transformed into neurophysiologically-motivated spectro-temporal input features. These patterns are called amplitude modulation spectrograms (AMS), as they contain information of both center frequencies and modulation frequencies within each 32 ms-analysis frame. The different representations of speech and noise in AMS patterns are detected by a neural network, which estimates the present SNR in each frequency channel. Quantitative experiments show a reliable estimation of the SNR for most types of nonspeech background noise. For noise suppression, the frequency bands are attenuated according to the estimated present SNR using a Wiener filter approach. Objective speech quality measures, informal listening tests, and the results of automatic speech recognition experiments indicate a substantial benefit from AMS-based noise suppression, in comparison to unprocessed noisy speech.

A speech pause detection algorithm is an important and sensitive part of most single-microphone noise reduction schemes for enhancement of speech signals corrupted by additive noise as an estimate of the background noise is usually determined when speech is absent. An algorithm is proposed which detects speech pauses by adaptively tracking minima in a noisy signal's power envelope both for the broadband signal and for the high-pass and low-pass filtered signal. In poor signal-to-noise ratios (SNRs), the proposed algorithm maintains a low false-alarm rate in the detection of speech pauses while the standardized algorithm of ITU G.729 shows an increasing false-alarm rate in unfavorable situations. These characteristics are found with different types of noise and indicate that the proposed algorithm is better suited to be used for noise estimation in noise reduction algorithms, as speech deterioration may thus be kept at a low level. It is shown that in connection with the Ephraim-Malah (1984) noise reduction scheme, the speech pause detection performance can even be further increased by using the noise-reduced signal instead of the noisy signal as input for the speech pause decision unit.

A novel noise suppression scheme for speech signals is proposed which is based on a neurophysiologically-motivated estimation of the local signal-to-noise ratio (SNR) in different frequency chan(cid:173) nels. For SNR-estimation, the input signal is transformed into so-called Amplitude Modulation Spectrograms (AMS), which rep(cid:173) resent both spectral and temporal characteristics of the respective analysis frame, and which imitate the representation of modula(cid:173) tion frequencies in higher stages of the mammalian auditory sys(cid:173) tem. A neural network is used to analyse AMS patterns generated from noisy speech and estimates the local SNR. Noise suppres(cid:173) sion is achieved by attenuating frequency channels according to their SNR. The noise suppression algorithm is evaluated in speaker(cid:173) independent digit recognition experiments and compared to noise suppression by Spectral Subtraction.

This paper describes the application of a quantitative psychoacoustical signal preprocessing model for objective speech quality measurement. The preprocessing is applied to transform the original and the distorted speech signal to an internal representation which is thought of as the information that is accessible to higher neural stages of perception. From a comparison of these internal representations a quality measure can be derived that shows a high correlation to the subjective MOS data of various test databases. The inherent parameters of the preprocessing model were derived directly from psychoacoustical data independent of the present study. The detection thresholds of codec-like distortions obtained in a psychoacoustical experiment could also be predicted by the model. This indicates that the internal representation contains the relevant information for detecting perceivable differences. It provides evidence for a direct relation between speech quality and detectability of a distortion.

The combination of a model of auditory perception (PEMO) as feature extractor and of a Locally Recurrent Neural Network (LRNN) as classifier yields promising ASR results in noise. Our study focuses on the interplay between both techniques and their ability to complement each other in the task of robust speech recognition. We performed recognition experiments with modifications of PEMO processing concerning amplitude compression and envelope modulation filtering. The results show that the distinct and sparse peaks of PEMO speech representation which are well maintained in noise are sufficient cues for LRNN-based recognition due to LRNN's ability to exploit information which is distributed over time. Enhanced envelope modulation bandpass filtering of PEMO feature vectors better reflects the average modulation spectrum of speech and further decreases the influence of noise. 1. INTRODUCTION One major problem in automatic speech recognition (ASR) is the robustness of ASR systems against no...

The present w ork concerns a system aimed at enhancing a target talker under varying signal conditions based on the use of several dierent t ypes of information or \cues".Toward this end, an architecture designed to combine separately operating estimators is described and evaluated.The architecture is currently implemented using spatial-and periodicitybased enhancement algorithms, and evaluated using a male target talker and female jammer talker under several spatial and target-to-jammer ratio (TJR) conditions.Using a TJR estimation algorithm, the implementation is shown to yield improved TJR under all tested input TJRs (-4, 0, 4, and 8 dB) and spatial conditions (target and jammer straight ahead; target ahead and jammer at 60 degrees).Improvement ranges from 1.4 to 4.5 dB.