Dynamic Cepstrum Research Articles

Static and Dynamic Spectral Features: Their Noise Robustness and Optimal Weights for ASR

In this paper, we investigate the relative noise robustness of dynamic and static spectral features in speech recognition. It is found that the dynamic cepstrum is more robust to additive noise than its static counterpart. The results are consistent across different types of noise and over a wide range of noise levels. To exploit this unequal robustness, we propose a simple yet effective strategy of exponentially weighting the likelihoods that are contributed by the static and dynamic features during the decoding process. The optimal weights are discriminatively trained with a small amount of development data. This method is evaluated on two speaker-independent, connected digit databases, one in English (Aurora 2) and the other in Cantonese (CUDIGIT). For various types of noise at different signal-to-noise ratios (SNRs), the average relative word error rate reductions attained with the discriminatively trained weights are 36.6% and 41.9% for Aurora 2 and CUDIGIT, respectively. Noticeable performance improvement can be observed even when there is channel distortion. The proposed approach is appealing to practical applications because 1) noise estimation is not required, 2) model adaptation is not required, 3)only a minor modification of the decoding process is needed, and 4) only a few feature weights need to be trained

Speech recognition under noisy environments using segmental unit input HMM

AbstractIn this paper, we apply a segmental unit input HMM to noisy speech recognition. In this modeling, several successive frames are combined and treated as an input vector. We expect that the segmental unit input HMM will be effective for noisy speech recognition because segmental statistics considering correlation between frames reduce noise effects when the correlation of noise between frames is assumed to be small. In recognition experiments, we compared the segmental unit input HMM with a conventional frame‐based HMM and found the segmental unit input HMM to be superior. We also compared the segmental unit input HMM with dynamic cepstral coefficients, which have both static and dynamic features, and found that the segmental unit input HMM is more effective than the dynamic cepstrum. We also combined the segmental unit input HMM with a spectral subtraction method and confirmed the effectiveness of the method. Additionally, in experiments using acoustic models trained with noisy speech, the segmental unit input HMM outperformed the conventional HMM. From these results, we propose PMC for the segmental unit input HMM. Experimental results showed the PMC for segmental unit input HMM offered better recognition performance than the original PMC. © 2002 Wiley Periodicals, Inc. Syst Comp Jpn, 33(8): 111–120, 2002; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/scj.1151

Channel normalization techniques for automatic speech recognition over the telephone

In this paper we aim to identify the underlying causes that can explain the performance of different channel normalization techniques. To this aim we compared four different channel normalization techniques within the context of connected digit recognition over telephone lines: cepstrum mean subtraction, the dynamic cepstrum representation, RASTA filtering and phase-corrected RASTA. We used context-dependent and context-independent hidden Markov models that were trained using a wide range of different model complexities. The results of our recognition experiments indicate that each channel normalization technique should preserve the modulation frequencies in the range between 2 and 16 Hz in the spectrum of the speech signals. At the same time, DC components in the modulation spectrum should be effectively removed. With context-independent models the channel normalization filter should have a flat phase response. Finally, for our connected digit recognition task it appeared that cepstrum mean subtraction and phase-corrected RASTA performed equally well for context-dependent and context-independent models when equal amounts of model parameters were used.

Cepstral representation of speech motivated by time-frequency masking: an application to speech recognition.

A new spectral representation incorporating time-frequency forward masking is proposed. This masked spectral representation is efficiently represented by a quefrency domain parameter called dynamic-cepstrum (DyC). Automatic speech recognition experiments have demonstrated that DyC powerfully improves performance in phoneme classification and phrase recognition. This new spectral representation simulates a perceived spectrum. It enhances formant transition, which provides relevant cues for phoneme perception, while suppressing temporally stationary spectral properties, such as the effect of microphone frequency characteristics or the speaker-dependent time-invariant spectral feature. These features are advantageous for speaker-independent speech recognition. DyC can efficiently represent both the instantaneous and transitional aspects of a running spectrum with a vector of the same size as a conventional cepstrum. DyC is calculated from a cepstrum time sequence using a matrix lifter. Each column vector of the matrix lifter performs spectral smoothing. Smoothing characteristics are a function of the time interval between a masker and a signal. DyC outperformed a conventional cepstrum parameter obtained through linear predictive coding (LPC) analysis for both phoneme classification and phrase recognition by using hidden Markov models (HMMs). Compared with speaker-dependent recognition, an even greater improvement over the cepstrum parameter was found in speaker-independent speech recognition. Furthermore, DyC with only 16 coefficients exhibited higher speech recognition performance than a combination of the cepstrum and a delta-cepstrum with 32 coefficients for the classification experiment of phonemes contaminated by noises.

Speech recognition method using time-frequency masking mechanism

A speech recognition method in which input speech signals are converted to digital signals and then time sequentially converted to cepstrum coefficients or logarithmic spectra. Dynamic spectrum time sequence is obtained by time frequency filtering of cepstrum coefficients, or masked spectrum time sequence is obtained by time frequency masking of the logarithmic vector time sequence. Based on the dynamic cepstrum time sequence or masked spectrum time sequence obtained in this manner, speech is recognized.

On the effectiveness of dynamic cepstrum in measuring the English /r/-/l/ acoustic-perceptual dissimilarity.

On the effectiveness of dynamic cepstrum in measuring the English /r/-/l/ acoustic-perceptual dissimilarity.

A new masked spectrum representation applied to English /r/–/l/ dissimilarity measurement

This paper proves that a dynamic cepstrum is effective not only in automatic speech recognition, but also in explaining the speech perception mechanism. Talker dependency on the perception of American English /r/–/l/ in Japanese listeners has been reported [J. S. Logan et al., J. Acoust. Soc. Am. 89, 874 (1991)]. In order to interpret this phenomenon with regard to the acoustical properties of the stimuli, talker dependency is assumed to be caused by the acoustical dissimilarity of the stimuli. This paper applies a dynamic cepstrum to measuring the dissimilarity of the stimuli. The dynamic cepstrum is a new spectral representation for automatic speech recognition that incorporates the time‐frequency characteristics of forward masking [K. Aikawa et al., J. Acoust. Soc. Am. 92, 2476(A) (1992)]. This parameter enhances formant shifts and suppresses stationary spectral features. In a perception experiment, identification tests for English /r/–/l/ minimal pairs uttered by five talkers were conducted. The dynamic cepstrum and a conventional cepstrum were compared for their performance in measuring the acoustical dissimilarity of the minimal pairs. The /r/–/l/ dissimilarity measured by the dynamic cepstrum shows a talker dependency highly correlated with the talker dependency of the perception experimental results. [Speech data were provided by Dr. D. B. Pisoni, Indiana University.]

Dynamic cepstral parameter incorporating time-frequency masking and its application to speech recognition

A ‘‘dynamic cepstrum’’ is proposed as a new spectral parameter that outperforms conventional cepstrum in speech recognition. The dynamic cepstrum incorporates both the static and dynamic aspects of speech spectral sequences by implementing forward masking, which is one of the most important mechanisms for extracting the spectral dynamics that provide acoustic cues in speech perception. Recent research on auditory perception [E. Miyasaka, J. Acoust. Soc. Jpn. 39, 614–623 (1983) (in Japanese)] reports that forward masking becomes more widespread over the frequency axis as the masker-signal interval increases. This masking characteristic facilitates the novel filtering methodology of time-dependent spectral filtering. The new dynamic cepstrum spectral parameter can simulate this function. The parameter is obtained by subtracting the masking level from the current cepstrum. The masking level at the current time is calculated as the sum of the masking levels obtained by filtering the preceding spectral sequence, where the cut-off frequency of the low-pass filter shifts lower as a function of masker-signal interval. A /b,d,g,m,n,N/recognition experiment that applies the dynamic cepstrum to hidden Markov models demonstrates that the dynamic cepstrum outperforms the combination of cepstrum and delta cepstrum, where the combination requires twice as many parameters as the dynamic cepstrum.