Previously, a dereverberation method based on generalized spectral subtraction (GSS) using multi-channel least mean-squares (MCLMS) has been proposed. The results of speech recognition experiments showed that this method achieved a significant improvement over conventional methods. In this paper, we apply this method to distant-talking (far-field) speaker recognition. However, for far-field speech, the GSS-based dereverberation method using clean speech models degrades the speaker recognition performance. This may be because GSS-based dereverberation causes some distortion between clean speech and dereverberant speech. In this paper, we address this problem by training speaker models using dereverberant speech obtained by suppressing reverberation from arbitrary artificial reverberant speech. Furthermore, we propose an efficient computational method for a combination of the likelihood of dereverberant speech using multiple compensation parameter sets. This addresses the problem of determining optimal compensation parameters for GSS. We report the results of a speaker recognition experiment performed on large-scale far-field speech with different reverberant environments to the training environments. The proposed GSS-based dereverberation method achieves a recognition rate of 92.2%, which compares well with conventional cepstral mean normalization with delay-and-sum beamforming using a clean speech model (49.0%) and a reverberant speech model (88.4%). We also compare the proposed method with another dereverberation technique, multi-step linear prediction-based spectral subtraction (MSLP-GSS). The proposed method achieves a better recognition rate than the 90.6% of MSLP-GSS. The use of multiple compensation parameters further improves the speech recognition performance, giving our approach a recognition rate of 93.6%. We implement this method in a real environment using the optimal compensation parameters estimated from an artificial environment. The results show a recognition rate of 87.8% compared with 72.5% for delay-and-sum beamforming using a reverberant speech model.

/pdf/distant-talking-speaker-identification-by-generalized-lfsc6nshvt.pdf

Distant-talking speaker identification by generalized spectral subtraction-based dereverberation and its efficient computation

In this paper, we propose a robust distant-talking speech recognition system with asynchronous speech recording. This is implemented by combining denoising autoencoder-based cepstral-domain dereverberation, automatic asynchronous speech (microphone or mobile terminal) selection and environment adaptation. Although applications using mobile terminals have attracted increasing attention, there are few studies that focus on distant-talking speech recognition with asynchronous mobile terminals. For the system proposed in this paper, after applying a denoising autoencoder in the cepstral domain of speech to suppress reverberation and performing Large Vocabulary Continuous Speech Recognition (LVCSR), we adopted automatic asynchronous mobile terminal selection and environment adaptation using speech segments from optimal mobile terminals. The proposed method was evaluated using a reverberant WSJCAMO corpus, which was emitted by a loudspeaker and recorded in a meeting room with multiple speakers by far-field multiple mobile terminals. By integrating a cepstral-domain denoising autoencoder and automatic mobile terminal selection with environment adaptation, the average Word Error Rate (WER) was reduced from 51.8% of the baseline system to 28.8%, i.e., the relative error reduction rate was 44.4% when using multi-condition acoustic models.

http://www.apsipa.org/proceedings_2014/Data/paper/1063.pdf

Denoising autoencoder and environment adaptation for distant-talking speech recognition with asynchronous speech recording

In this paper, we propose a robust distant-talking speech recognition system with asynchronous speech recording. This is implemented by combining automatic asynchronous speech (microphone or mobile terminal) selection and environmental adaptation with deep neural network based framework. Although applications using mobile terminals have attracted increasing attention, there are few studies that focus on distant-talking speech recognition with asynchronous mobile terminals. For the system proposed in this paper, by using bottleneck Features (BFs) from a Deep Neural Network (DNN) rather than the conventional Mel-Frequency Cesptral Coefficients (MFCCs), we adopted the state-of-the-art deep neural network acoustic model, environmental adaptation and automatic asynchronous speech selection. The proposed method was evaluated by using a reverberant WSJCAM0 corpus, which was emitted by a loudspeaker and recorded in a meeting room with multiple speakers by far-field multiple mobile terminals. By using the bottleneck features based DNN acoustic model with automatic asynchronous speech selection and environmental adaptation, the average Word Error Rate (WER) was reduced from 55.32% of the baseline system to 19.38%, i.e. the relative error reduction rate was 64.97%.

/pdf/speech-selection-and-environmental-adaptation-for-2au3tkin4k.pdf

Speech selection and environmental adaptation for asynchronous speech recognition

The impulse response of a reverberant environment, in general, is a nonminimum phase and cannot be inverted. But an exact inverse of the environment can be obtained by modeling the room as a multiple input–output (MINT) system [M. Miyoshi and Y. Kaneda ICASSP (1986)]. In this report, this model is applied to a microphone array and is used as a front‐end processor for a speaker verification system. The G matrix is inverted using row action projection (RAP), an iterative approach to solving a system of linear equations. Starting from an initial guess, the solution is repeatedly projected onto each hyperplane of the equation system until it converges. The method is stable, robust to noise, and converges to the pseudo‐inverse solution. In computer‐simulated experiments, the signal‐to‐reverberant‐noise ratio is found to improve with the number of microphones in the array. A speaker verification system using the array is evaluated at various signal‐to‐competing‐noise ratios (SCNR). Results suggest that verification performance can be substantially elevated in adverse acoustic environments.

/pdf/stable-dereverberation-using-microphone-arrays-for-speaker-3z2cevnvlf.pdf

J. L. Flanagan

Papers

Stable dereverberation using microphone arrays for speaker verification