Design of 3D microphone arrays using automatic speech recognition and psychoacoustic evaluation for immersive sounds

Abstract

A microphone array system was developed to capture the speech for the transmission to a remote space. The captured signals by the different types of arrays were synthesized into the binaural signals and evaluated for speech recognition. The automatic speech recognition algorithm, based on hidden Markov model, was applied to quantify the accuracy of speech recognition with a word list. The psychoacoustic experiments were performed to assess the quality of sound immersion. Several noise environments were considered with different signal to noise ratios and the multi-channel sound capturing systems was found as most efficient. ICS file for iCal / Outlook