Channel normalization techniques for automatic speech recognition over the telephone

We compared three different channel normalisation (CN) methods in the context of a connected digit recognition task over the phone: cepstrum mean substraction (CMS), RASTA filtering and the Gaussian dynamic cepstrum representation (GDCR). Using a small set of context independent (CI) continuous Gaussian mixture hidden Markov models (HMMs), we found that CMS and RASTA outperformed the GDCR technique. We show that the main cause for the superiority of CMS compared to RASTA is the phase distortion introduced by the RASTA filter. Recognition results for a phase corrected RASTA technique are identical to those of CMS. Our results indicate that an ideal cepstrum based CN method should: (1) effectively remove the DC component; (2) at least preserve modulation frequencies in the range 2-16 Hz; and (3) introduce no phase distortion in case CI HMMs are used for recognition.

/pdf/comparison-of-channel-normalisation-techniques-for-automatic-1i1dp1utd7.pdf

Comparison of channel normalisation techniques for automatic speech recognition over the phone

Saying that late 20th century automatic speech recognition (ASR) is pattern recognition, is something of a truism, but perhaps one of which the fundamental implications are not always fully appreciated. Essentially, a pattern recognition task boils down to measuring the distance between a physical representation of a new, as yet unknown token, and all elements of a set of preexisting patterns, of course in the same physical representation. On the one hand, the ‘patterns’ that can be recognized are, implicitly or explicitly, separate and invariable entities. For example, the command open in a Windows control application always has the same invariable and unique meaning. On the other hand, the unknown input tokens are continuous signals that typically show a high degree of variability. ASR research has centered around the problem of how to map continuous, variable acoustic representations onto discrete, invariable patterns. In ASR the physical representation of the speech tokens is some kind of dynamic power spectrum, for reasons which date back to the days of Ohm and von Helmholtz, who have shown that the power spectrum explains most of the perceptual phenomena in human speech processing. Since the inception of digital signal processing dynamic spectra are approximated by a sequence of short-time spectra (Rabiner & Schafer 1978). Consequently, the pattern match in ASR is invariably implemented as the accumulation of some distance measure between the acoustic features derived from a sequence of short-time spectra of the input token and the corresponding representation of the active patterns (see Fig. 2.1). Therefore, anything which adds to the variability of the short time spectrum of a speech signal will, as it were by definition, complicate pattern matching, and consequently complicate ASR.

Acoustic Features and Distance Measure to Reduce Vulnerability of ASR Performance Due to the Presence of a Communication Channel and/or Background Noise

A method and apparatus for removing convolution noise from a signal such as one carrying speech information The signal is transformed into a log-spectral domain where a smoothed model is fitted to the log-spectrum subject to constraints of concavity and an overall bandpass shape The smoothed model has quadratic segments of negative curvature and linear segments, the segments being smoothly joined at breakpoints The model, which may be recursively updated, is substracted from each log-spectral data vector

Accelerated convolution noise elimination

Comparison of channel normalisation techniques for automatic speech recognition over the telephone

Human auditory perception is perfectly capable to deal with time-invariant linear filter effects, such as those introduced by telephone handsets and telephone channels. We compared two different schemes for modeling human auditory time-frequency masking: RASTA filtering and the dynamic cepstrum represen­ tation (DCR). We used a small set of context-independent phone hidden Markov models for a recognition task of connected digit strings over the telephone. We found that RASTA filtering out­ performed the Gaussian DCR approach, despite the fact that RASTA represents a more crude approximation of human for­ ward masking. Our results may be influenced by the choice of the mel-frequency cepstral representation that we used. The superiour performance of the RASTA technique may also be ex­ plained by the fact that the frequency response of the RASTA filter is better matched to the region of modulation frequencies where human auditory perception is most sensitive.

J.M. de Veth

Papers

Channel normalisation by using rasta filtering and the dynamic cepstrum for automatic speech recognition over the phone