Effect of spectral normalization on different talker speech recognition by cochlear implant users

Effects of Within-Talker Variability on Speech Intelligibility in Mandarin-Speaking Adult and Pediatric Cochlear Implant Patients

Cochlear implant (CI) speech performance is typically evaluated using well-enunciated speech produced at a normal rate by a single talker. CI users often have greater difficulty with variations in speech production encountered in everyday listening. Within a single talker, speaking rate, amplitude, duration, and voice pitch information may be quite variable, depending on the production context. The coarse spectral resolution afforded by the CI limits perception of voice pitch, which is an important cue for speech prosody and for tonal languages such as Mandarin Chinese. In this study, sentence recognition from the Mandarin speech perception database was measured in adult and pediatric Mandarin-speaking CI listeners for a variety of speaking styles: voiced speech produced at slow, normal, and fast speaking rates; whispered speech; voiced emotional speech; and voiced shouted speech. Recognition of Mandarin Hearing in Noise Test sentences was also measured. Results showed that performance was significantly poorer with whispered speech relative to the other speaking styles and that performance was significantly better with slow speech than with fast or emotional speech. Results also showed that adult and pediatric performance was significantly poorer with Mandarin Hearing in Noise Test than with Mandarin speech perception sentences at the normal rate. The results suggest that adult and pediatric Mandarin-speaking CI patients are highly susceptible to whispered speech, due to the lack of lexically important voice pitch cues and perhaps other qualities associated with whispered speech. The results also suggest that test materials may contribute to differences in performance observed between adult and pediatric CI users.

Perception of speech produced by native and nonnative talkers by listeners with normal hearing and listeners with cochlear implants

PURPOSE The aim of this study was to evaluate the understanding of English sentences produced by native (English) and nonnative (Spanish) talkers by listeners with normal hearing (NH) and listeners with cochlear implants (CIs). METHOD Sentence recognition in noise was measured in adult subjects with CIs and subjects with NH, all of whom were native talkers of American English. Test sentences were from the Hearing in Noise Test (HINT) database and were produced in English by four native and eight nonnative talkers. Subjects also rated the intelligibility and accent for each talker. RESULTS The speech recognition thresholds in noise of subjects with CIs and subjects with NH were 4.23 dB and 1.32 dB poorer with nonnative talkers than with native talkers, respectively. Performance was significantly correlated with talker intelligibility and accent ratings for subjects with CIs but only correlated with talker intelligibility ratings for subjects with NH. For all subjects, performance with individual nonnative talkers was significantly correlated with talkers' number of years of residence in the United States. CONCLUSION CI users exhibited a larger deficit in speech understanding with nonnative talkers than did subjects with NH, relative to native talkers. Nonnative talkers' experience with native culture contributed strongly to speech understanding in noise, intelligibility ratings, and accent ratings of both listeners with NH and listeners with CIs.

Voice conversion in cochlear implantation

OBJECTIVES/HYPOTHESIS

Voice conversion algorithms may benefit cochlear implant (CI) users who better understand speech produced by one talker than by another. It is unclear how the source or target talker's fundamental frequency (F0) information may contribute to perception of converted speech. This study evaluated voice conversion algorithms for CI users in which the source or target talker's F0 was included in the converted speech.

STUDY DESIGN

Development and evaluation of computerized voice conversion algorithms in CI patients.

METHODS

A series of cepstral analysis-based algorithms were developed and evaluated in six CI users. The algorithms converted talker voice gender (male-to-female, or female-to-male); either the source or target talker F0 was included in the converted speech. The voice conversion algorithms were evaluated in terms of recognition of IEEE sentences, speech quality, and voice gender discrimination.

RESULTS

Voice gender recognition performance showed that listeners strongly cued to the F0 that was included within the converted speech. For both IEEE sentence recognition and voice quality ratings, performance was poorer with the voice conversion algorithms than with original speech. Performance on female-to-male conversion was superior to male-to-female conversion.

CONCLUSION

The strong cueing to F0 within the voice conversion algorithms suggests that CI users are able to utilize temporal periodicity information for some pitch-related tasks. Limitations on spectral channel information experienced by CI users may result in poorer performance with voice conversion algorithms due to distortion of speech formant information and degradation of the spectral envelope.

Effects of speaking style on speech intelligibility for Mandarin-speaking cochlear implant users

Cochlear implant (CI) users' speech understanding may be influenced by different speaking styles. In this study, speech recognition was measured in Mandarin-speaking CI and normal-hearing (NH) subjects for sentences produced according to four styles: slow, normal, fast, and whispered. CI subjects were tested using their clinical processors; NH subjects were tested while listening to a four-channel CI simulation. Performance gradually worsened with increasing speaking rate and was much poorer with whispered speech. CI performance was generally similar to NH performance with the four-channel simulation. Results suggest that some speaking styles, especially whispering, may negatively affect Mandarin-speaking CI users' speech understanding.

Are changes in pitch and formants enough to influence talker normalization processes in children and adults?

In the present study, we examined the influence of mean F0 and formant values on talker normalization. Initially, two speakers recorded an identical set of 10 isophonemic word lists in Hebrew, consisting of 10 words each. These recordings were then manipulated by means that affect F0 only, or both F0 and formant frequencies. Different degrees of manipulation were carried out in order to create a new set of lists which were perceived to have been obtained from multiple talkers. The original and manipulated lists were presented to a group of 12 adults and 12 children at 30 dB above pure-tone average thresholds (PTA). Half of the listeners was presented with the single-talker lists first and then the multiple-talker lists, and the other half was presented with the word lists in reversed order. Listeners were instructed to repeat the words they heard. Correct word recognition was scored. The results indicated: (1) lower word recognition scores for the multiple-talker lists than for the single-talker lists, for both children and adults, (2) larger inter-subject variability in performance for the multiple-talker compared to the single-talker lists for both age groups, and (3) that order of presentation influenced word recognition of the multiple-talker lists only, being worse when presented first. Our findings support the following: (1) manipulations of F0 and formants of a single talker are sufficient to influence talker normalization processes, (2) this influence varies between listeners suggesting that listeners do not use the same acoustic information in their speech perception process, and (3) even adults, who are proficient in the language, do not eliminate entirely the irrelevant talker-specific acoustic information in the speech perception process.