Shigeyoshi Kitazawa

Discrimination of isolated vowels and stop consonants using features extracted from onset spectrum

Place and/or manner of articulation of the 7 consonants (/?,p,t,k,b,d,g/) are discriminated in an enviroment of /a,i,u, e,o/. We considered that isolated vowels are approximately characteriszed as the glottal stop /?/ and included them as stop consonants. The feature vector used was spectrum of the initial 50ms after vowel voicing or stop burst point. The burst point of time was manually specified referring to waveform display. We used a stepwise discriminant analysis program with variable selection to classify the samples of speech. The sample set comprised 3402 samples from 89 male speakers. Some invariance was found in the discrimination between /?,p,t,k/ achieving 89% correct score across vowel context and speakers. This is a 3% degradation from that between /p,t,k/. A few samples of voiced stops lacking murmur preceding the burst point was well discriminated from voiceless stops of same point of articulation.

Comparison of two signal processing strategies using consonant recognition tests

Conclusion We have experienced 35 cochlear implantations with and without EPS. Fortunately, there were no severe complications, and there were no significant differences between electrode impedance values with and without EPS. The Clarion device has a pre-curved electrode in itself, which may reduce the electrode impedance by hugging the cochlear modiolus. However, more data and long-standing follow-up is needed to evaluate the language development and hearing outcomes without EPS.

Speaker‐independent consonant recognition by integrating discriminant analysis and hmm

A new consonant recognition method is proposed which integrates two stochastic methods: discriminant analysis and HMM. Discriminant analysis is effective in analyzing local patterns but it assumes precise detection of reference points. HMM is able to extract the overall dynamic features and needs no explicit segmentation of speech. However, it lacks the ability to discriminate between similar consonants. The method herein constructs HMM with discriminant analysis front end and recognizes consonants by combining the score obtained by discriminant analysis and that by HMM. For all the Japanese consonants, this integrated method achieved the recognition rate of 92.1 percent, which is higher by 5 to 15 percent than the case using either of the two methods alone.

Study of effect of low‐pass filtering for channels amplitude estimation on speech intelligibility by acoustic simulation of SPEAK strategy

The conventional way of amplitude envelope estimation for stimulation pulse train generation for selected channels within multi‐channel cochlear implant speech processors is low‐pass filtering of half or full wave rectified (HWR or FWR) frequency bands of the corresponding channels. To study the necessity of low‐pass filtering (LPF) for amplitude envelope generation, acoustic simulation of the SPEAK speech strategy was implemented. Stimulation pulse patterns for selected channels are estimated without LPF and compared to those estimated by the conventional way. Acoustic stimuli are regenerated using an overlap adding technique of filtered pulse trains. Speech intelligibility tests on CV and VCV of Japanese within normal‐hearing subjects were conducted to measure the effect of LPF. A one‐way ANOVA indicates that the LPF with a cutoff frequency 160 Hz for amplitude envelope estimation of selected channels negatively affected the speech intelligibility.

TEMAS: Visualization of temporal variation in spontaneous speech

This study presents a new method for analyzing speech rhythm. First, speech speed is measured and displayed on the basis of syllable (or mora) and stress (rhythmic foot). Then, this specific algorithm TEMAX (temporal evaluation and measurement algorithm by KS) is applied to the speech envelope sampled at 40 Hz; speech wave is half‐wave rectified and low‐pass filtered at 20 Hz. The DFT of the envelope using a 1‐s window is convenient to set off isosyllabic characteristics. For Japanese, the TEMAX‐gram, a sonagraphic output, traces two dark bars, called rhythmic formants: RF1 and RF2: the first one, around 8 Hz, and the second one, at about halfway. RF1 corresponds to speech rate, which appears almost steady in read speech and monolog but shows wide variations in spontaneous speech. RF2 represents the bimoraic rhythmic foot, that is, a combination of two adjacent moras forming a single large power peak. Considering English, its isochronic characteristics are observable with a 2‐s window as RF1. Furthermore,...