Ambiguity in phoneme recognition when using a bone conduction microphone
Karl Buck, Vronique Zimpfer · The Journal of the Acoustical Society of America · 2011
Bone conduction microphones are often promoted as good means for recoding voice signals in very noisy and/or harsh environments. However, it is reported in the literature that the speech intelligibility when using contact microphones is, in all noise conditions, worse than when using a boom microphone. Therefore, long term spectra of speech signals simultaneously recorded with bone and air microphones have been analyzed. This analysis shows, for all measured subjects, that the module of the transfer function between the signal recorded at the air-microphone and, because of this, at the contact microphone increases toward higher frequencies. The results also show large interindividual differences. In order to get more perception related information, a corpus of French vowels has been recorded simultaneously with air and contact microphone. This corpus has been used for a listening test aiming to show confusion between different vowels. The main confusions are for vowels having the same frequency of the first formant (e.g., [i][y][u]) which then are perceived as the central vowel. These confusions could explain the systematically lower intelligibility reported for speech when recorded with contact microphones. Planned experiments using plosive and fricative consonants should give more information about transfer mechanisms of bone conducted voice.