Recognition of Noisy Speech: Using Minimum-Mean Log-Spectral Distance Estimation

Recognition of Noisy Speech: Using Minimum-Mean Log-Spectral Distance Estimation A. Ere11 and M. Weintraub SRI International 333 Ravenswood Ave. Menlo Park, CA 94025 Abstract A model-based spectral estimation algorithm is derived that improves the robustness of speech recognition systems to additive noise. The algorithm is tailored for filter-bank-based systems, where the estimation should seek to minimize the distortion as measured by the recognizer's distance metric. This estimation criterion is approximated by minimizing the Euclidean distance between spectral log-energy vectors, which is equivalent to minimizing the nonweighted, nontruncated cepstral distance. Correlations between frequency channels are incorporated in the estimation by modeling the spectral distribution of speech as a mixture of components, each representing a different speech class, and assuming that spectral energies at different frequency channels are uncorrelated within each class. The algorithm was tested with SRI's continuous-speech, speaker-independent, hidden Markov model recognition system using the large- vocabulary NIST "Resource Management Task." When trained on a clean-speech database and tested with additive white Gaussian noise, the new algorithm has an error rate half of that with MMSE estimation of log spectral energies at individual frequency channels, and it achieves a level similar to that with the ideal condition of training and testing at constant SNR. The algorithm is also very efficient with additive environmental noise, recorded with a desktop microphone. I. Introduction Speech-recognition systems are very sensitive to differences between the testing and training conditions. In particular, systems that are trained on high-quality speech degrade drastically in noisy environments. Several methods for handling this problem are in common use, among them supplementing the acoustic front end of the recognizer with a statistical estimator. This paper introduces a novel estimation algorithm for a filter-bank-based front end and describes recognition experiments with noisy speech. The problem of designing a statistical estimator for speech recognition is that of defining an optimality criterion that will match the recognizer, and deriving an algorithm to compute the estimator based on this criterion. Defining the optimal criterion is easier for speech recognition than it is for speech enhancement for human listeners, since the signal processing is known in the former, but not in the latter. For a recognition system that is based on a distance metric, whether for template matching or vector quantization, a reasonable criterion would be to minimize the average distortion as measured by the distance metric. In practice, achieving this criterion may turn out not to be feasible, and the question is then to what extent the computationally feasible methods approximate the desired optirnality criterion. A basic difference between the cepstral distance criterion and the MMSE of single frequency channels (whether DFT coefficients or filter energies) is that the former implies a joint estimate of a feature vector, whereas the latter implies an independent estimation of scalar variables. Because the speech spectral energies at different frequencies are correlated, an independent estimate of individual channels results in a suboptimal estimation. To incorporate part of the correlations in the estimator, we modified our single-channel MMSE to be …

Content truncated to 500 words