荒井 隆行

This paper describes digital signal processing (DSP) microprocessor experiments designed for university juniors majoring in electric & electronic (E gr E) engineering. At the time of enrollment, most students have only studied the curriculum for analog signal processing which is taken in the prior semester. The proposed DSP microprocessor experiments are included along with analog signal processing. This early-on introduction to DSP technology allows students to realize that, in comparison with analog circuits, DSP microprocessors can process the same signals in real-time with broader flexibility. Such an understanding is considered important to instill strong incentive fur students to become interested in the field of DSP.

In this paper we investigate how polarity inversion of speech signals effects human perception, and we apply this technique for data hiding. In most languages, glottal airflow during phonation is uni-directional, causing constant polarity of the speech waveform. On the other hand, the human auditory system cannot discriminate between speech signals with positive and negative polarity. Based on these facts, we developed an algorithm to hide data in speech signals. We assigned one bit to each syllable of speech, and inverted the polarity of the signal at every syllable according to the assigned bit. We performed a test using 20 sentences from the TIMIT corpus to determine both whether a human could distinguish between the original and polarity-inverted signal and whether we could automatically restore the embedded binary data. We found that we were able to successfully hide data and restore it automatically.

In this paper we report on a meth od for reducing the degradation of speech intelligibility in public halls caused severe reverberation. Hall reverberation makes speech more difficult to understand, particularly for the hearing-impaired. Our method involves processing the speech audio signal between a microphone and a loudspeaker that radiates the speech into the room. As there is a strong correlation between the modulation spectrum and the intelligibility of speech, we filtered the speech in the modulation frequency domain. Using several modulation filters, we conducted perceptual experiments with hearing-impaired subjects and asked their preference in a church. The experiments indicate that enhancing the modulation frequencies between 2 and 8 Hz improves intelligibility in reverberant environments. The four hearing-impaired subjects rated the processed speech easier to hear than the unprocessed speech.

We measured the accuracy of speech recognition as a function of band-pass filtering of the time trajectories of spectral envelopes. We examined (i) several types of recognizers such as dynamic time warping (DTW) and hidden Markov model (HMM), and (ii) several types of features, such as filter bank output, mel-frequency cepstral coefficients (MFCC), and perceptual linear predictive (PLP) coefficients. We used the resulting recognition data to determine the relative importance of information in different modulation spectral components of speech for automatic speech recognition. We concluded that: (1) most of the useful linguistic information is in modulation frequency components from the range between 1 and 16 Hz, with the dominant component at around 4 Hz; (2) in some realistic environments, the use of components from the range below 2 Hz or above 16 Hz can degrade the recognition accuracy. (C) 1999 Elsevier Science B.V. All rights reserved.

The intelligibility of syllables whose cepstral trajectories were temporally filtered was measured. The speech signals were transformed to their LPC cepstral coefficients, and these coefficients were passed through different filters. These filtered trajectories were recombined with the residuals and the speech signal reconstructed. The intelligibility of the reconstructed speech segments was then measured in two perceptual experiments for Japanese syllables. The effect of various low-pass, high-pass, and bandpass filtering is reported, and the results summarized using a theoretical approach based on the independence of the contributions in different modulation bands. The overall results suggest that speech intelligibility is not severely impaired as long as the filtered spectral components have a rate of change between 1 and 16 Hz. (C) 1999 Acoustical Society of America. [S0001-4966(99)01705-1].

We report on the effect of band-pass filtering of the time trajectories of spectral envelopes on speech recognition. Several types of filter (linear-phase FIR, DCT, and DFT) are studied. Results indicate the relative importance of different components of the modulation spectrum of speech for ASR. General conclusions are: (1) most of the useful linguistic information is in modulation frequency components from the range between 1 and 16 Hz, with the dominant component at around 4 Hz, (2) it is important to preserve the phase information in modulation frequency domain, (3) The features which include components at around 4 Hz in modulation spectrum outperform the conventional delta features, (4) The features which represent the several modulation frequency bands with appropriate center frequency and band width increase recognition performance.

The spectrum of spoken sentences was partitioned into quarter-octave channels and the onset of each channel shifted in time relative to the others so as to desynchronize spectral information across the frequency axis. Human listeners are remarkably tolerant of cross-channel spectral asynchrony induced in this fashion. Speech intelligibility remains relatively unimpaired until the average asynchrony spans three or more phonetic segments. Such perceptual robustness is correlated with the magnitude of the low-frequency (3-6 Hz) modulation spectrum and thus highlights the importance of syllabic segmentation and analysis for robust processing of spoken language. High-frequency channels (>1.5 kHz) play a particularly important role when the spectral asynchrony is sufficiently large as to significantly reduce the power in the low-frequency modulation spectrum (analogous to acoustic reverberation) and may thereby account for the deterioration of speech intelligibility among the hearing impaired under conditions of acoustic interference (such as background noise and reverberation) characteristic of the real world.

CMS法や動的特徴は変調周波数特性を操作することにより音声認識性能が向上することが知られているが、どの変調周波数がどの程度重要であるのかという定量的な検討は行われていない。そこで本研究では、様々な変調周波数特性を持った入力に対し、音声認識性能の違いを種々の雑音環境、認識方式、特徴量のもとで調べた。その結果、以下のことが分かった: 1)言語情報のほとんどが1〜16Hzの変調周波数帯域に存在し、その中でも4Hz付近が最も重要である。2)変調スペクトルにおいては位相情報も重要である。3)4Hz付近の変調周波数を含む特徴量を用いることで動的特徴量と同等以上の結果が得られる。4)適切な中心周波数と帯域幅をもつ複数のサブバンドを変調周波数上で用いることにより、認識性能がさらに向上する。

Palatalized articulation (PA) is frequently observed in speech uttered by postoperative cleft palate patients. Provided the acoustical and perceptual cues of PA can be found, speech therapists will be able to use these cues to diagnose PA non-invasively and objectively. We tested human perception of certain synthetic sounds to verify the cues of the PA of /s/ in Japanese. To synthesize the fricatives, we modified the center frequency and the bandwidth of a complex-conjugate pole pair of an all-pole filter obtained from the linear predictive analysis of the PA of /s/. First, we shifted the center frequency from 1,000 to 3,000 Hz, while the relative bandwidth, or Q factor, was fixed at 10. Subsequently, we shifted the Q factor from 1 to 10, while the center frequency was fixed at 1,800 Hz. The results of a perceptual experiment involving nine speech therapists were conclusive that fricatives having a peak between 1,600 and 2,400 Hz tend to be identified as the PA of /s/, and fricatives having a peak at 1,800 Hz with the Q factor &gt5, tend to be identified as the PA of /s/. The two-tube model also showed that a peak around 2 kHz characterizes the PA of /s/.

Our procedure of real-zero conversion uses a spectrum-reversal technique to convert the information of a bandlimited signal to real zeros, We conducted a simple reconstruction experiment and showed that our proposed method is essentially equivalent to the conventional technique of sine-wave crossings.

The effect of filtering the time trajectories of spectral envelopes on speech intelligibility was investigated. Since LPC cepstrum forms the basis of many automatic speech recognition systems, we filtered time trajectories of LPC cepstrum of speech sounds, and the modified speech was reconstructed after the filtering. For processing, we applied tow-pass, high-pass and band-pass filters. The results of the accuracy from the perceptual experiments for Japanese syllables show that speech intelligibility is not severely impaired as long as the filtered spectral components have 1) a rate of change faster than 1 Hz when high-pass filtered, 2) a rate of change slower than 24 Hz when low-pass filtered, and 3) a rate of change between 1 and 16 Hz when band-pass filtered.

In this paper approaches to language identification based on the sequential information of phonemes are described. These approaches assume that each language can be identified from its own phoneme structure, or phonotactics. To extract this phoneme structure, we use phoneme classifiers and grammars for each language. The phoneme classifier for each language is implemented as a multi-layer perceptron trained on quasi-phonetic hand-labeled transcriptions. After training the phoneme classifiers, the grammars for each language are calculated as a set of transition probabilities for each phoneme pair. Because of the interest in automatic language identification for worldwide voice communication, we decided to use telephone speech for this study. The data for this study were drawn from the OGI (Oregon Graduate Institute)-TS (telephone speech) corpus, a standard corpus for this type of research. To investigate the basic issues of this approach, two languages, Japanese and English, were selected. The language classification algorithms are based on Viterbi search constrained by a bigram grammar and by minimum and maximum durations. Using a phoneme classifier trained only on English phonemes, we achieved 81.1% accuracy. We achieved 79.3% accuracy using a phoneme classifier trained on Japanese phonemes. Using both the English and the Japanese phoneme classifiers together, we obtained our best result: 83.3%. Our results were comparable to those obtained by other methods such as that based on the hidden Markov model.