Objective estimation of tracheoesophageal speech ratings usingan auditory model

摘要

Total laryngectomy is often the treatment of choice for many individuals diagnosed with advancedlaryngeal cancer. This procedure alters the normal voice production mechanism, andtracheoesophageal (TE) speech is one alternative method of voicing postlaryngectomy. TE speech iscreated when pulmonary air is passed through the upper esophagus to create a vibratory source thatis then articulated into speech. TE speech is often characterized by abnormal voice quality. Acousticanalysis of TE speech has the potential of quantifying the voice quality and assisting the speechlanguage pathologist in facilitating rehabilitation. Motivated in part by the recent advances intelecommunication industry for speech quality estimation, this paper investigated the application ofan auditory model in predicting the ratings of TE speech by normal hearing listeners. The Moore-Glasberg auditory model was employed to extract perceptually relevant features from the acousticwaveform, and these features were later combined to estimate the subjective ratings of TE speech.This approach was validated with a database of subjective ratings of speech samples recorded from35 TE speakers. Results showed moderate correlations between the objective metrics and thesubjective ratings, and these correlations were significantly better than those obtained withtraditional methods used in the telecommunication applications.