Melody contour identification and instrument recognition using semitone mapping in Nucleus Cochlear Implant recipients

Abstract

Cochlear implants (CIs) were originally aimed at restoring speech perception for patients with profound hearing loss. Many postlingually deafened CI patients report that music is not well perceived while others enjoy it. Music consists of complex sounds composed of tones with harmonic structure of overtones and temporal fine structure. The harmonic structure is not preserved by the current standard (Std) ACE (advanced combination encoders) mapping and the temporal fine structure is not well encoded. The mapping is believed to produce distortion due to compression oin the low frequency range. In 2008 we proposed two new semitone (Smt) mappings (Smt-LF and Smt-MF) in two frequency ranges (130-1502 Hz and 440-5040 Hz) respectively (Omran et al. 2008). Smt mapping is expected to preserve the harmonic structure representation of overtones and this may improve melody recognition with CIs. In this paper two psychoacoustic experiments (melody contour identification (MCI) (Galvin et al. 2007) and instrument recognition (IR)) were conducted with three different conditions (Std, Smt-MF and Smt-LF mappings) with CI recipients by streaming processed stimuli directly to the implant. The MCI test included five patterns (rising - rising falling - flat - falling rising – falling). Each pattern consisted of five tones; each tone had a fundamental frequency and four overtones. The lowest fundamental frequency of each pattern is called “root”. Signals had two different roots A3 (220 Hz) and A4 (440 Hz). Proposed nine patterns with three roots (A3, A4 and A5) by Galvin et al. (2007) were examined in a pilot test. This test took a long time and the preliminary results showed a possibility to reduce the number of patterns to five and eliminate the fifth octave root (A5). In the IR test, four pairs of instruments (Trumpet and Trombone, Flute and Clarinet, Violin and Cello, Guitar and Piano) from four groups (Brass, Woodwind, Struck and String instruments) respectively were used. MCI and IR tests were conducted with 8 CI recipients. Results from MCI tests showed an improvement with Smt mapping in respect to Std mapping or at least similar results. However, wrong identification occurred with notes having filtered out partials. CI recipients showed an increase in identifying melody contour patterns with Smt mappings. Instrument identification performance decreased with semitone mappings.