An improved vocoder algorithm based on music harmonics and time sampling

Music plays an essential role in the healthy life of humans. However, few sound processor algorithms effectively encode the fine structure cues of music, resulting in inferior music perception for cochlear implant (CI) users, especially in pitch and melody. In this study, an improved music vocoder algorithm is proposed based on the conjunction of harmonic and time sampling (HTS). The algorithm includes two branches: the first, the pitch (i.e., fundamental frequency, F0) and important harmonics of music signal are extracted. Second, on the existing CI channels, the music signal is split into multiple sub-bands and the relevant envelopes are respectively matched to the F0 and important lower harmonics, then modulated and aligned with appropriate intervals which are not less than the auditory nerve response absolute refractory period (ANR-ARP). The violin sounds were synthesized and experimented with the CI tone vocoder, and the HTS algorithm was compared and evaluated with the classical continuous interleaved sampling (CIS) algorithm. Twenty normal hearing (NH) subjects were recruited for audiometry experiments. The results showed that the pitch ranking scores of the HTS were obviously better than that of the CIS, and in quiet and noisy conditions the melody recognition rates of the HTS were 46.4% and 49% higher than that of the CIS, respectively. And further results showed that the HTS algorithm also increased the timbre perception of CI vocoder music. It is suggested that the HTS algorithm has the potential to enhance the music perception of CI users. (C) 2023 Elsevier Ltd. All rights reserved.