| Popis: |
Enhanced spectral and temporal processing (STEP) was developed with the aim to improve the perception of indexical information carried by voice pitch, such as voice gender. A dual filter-bank approach was employed with a bank of narrow, good quality filters and a bank of parallel wide filters. The voice gender identification was compared with vowel pitch ranking task in 7 Nucleus CI recipients using two coding variants: ACE and STEP. The gender of the speaker was assessed in a 2-AFC task employing short 2-sec speech items extracted from 20 male and 20 female speakers. Vowel pitch ranking was evaluated in two frequency ranges, lower (LFR) in the 126 -164 Hz range and higher (HFR) in the 164- 212 Hz range, roughly corresponding to a male to female F0 continuum. Analysis of vowel pitch ranking indicated two of seven subjects having high performance with clear pattern for LFR and less clear pattern for HFR whether with ACE or STEP. One subject had very good performance in LFR, but in the HFR performance with ACE was clearer than for STEP. One subject had a clear pattern in LFR with STEP, but a poor pattern with ACE and a poor performance in the HFR with both variants. Preliminary data from regression bootstrap analysis indicates correlation tendency between cumulative d’ (dee-prime) and gender accuracy for ACE (r=0, 726 ; 95% CI=0, 128- 0, 988) and STEP (r=0, 651 ; 95% CI=0, 285-0, 993) coding variants in LFR. Same tendency is present between global d’ and gender accuracy for ACE (r=0, 739, 95% CI=0, 176-0, 988) and STEP (r=0, 675, 95% CI=0, 332-0, 985) in LFR. However, data analysis did not show any correlation tendency between either cumulative d’ or global d’ and gender accuracy in HFR. Voice gender identification has not improved in CI subjects using STEP as a new sound coding variant. Six of seven CI subjects had high performance in voice gender identification with higher score when detecting female voices (F0 > 164 Hz). Lack of correlation tendency in HFR and gender accuracy indicates that CI subjects were good in both tasks: vowel pitch and gender discrimination. In LFR, the results show a trend between successful vowel pitch and gender discrimination. CI subjects who were good in gender discrimination using ACE also had good performance using novel sound coding strategy – STEP. Generally it appears that improving temporal pitch coding with STEP had some effect in the LFR but no effect in the HFR. Funding This research was supported by Cochlear, and a part of the work by DK was funded by the European Commission (FP7- CIG- 2011-303927) |
