Application of Auditory Masking based Speech Denoising in Automotive Environments

Abstract

This paper presents an application experiment of denoising speech in the automotive field. An algorithm based on the auditory masking phenomenon was programmed and deployed for this purpose. Synthetic composite recordings of speech and vehice interior noise were used for three types of vehicles equipped with internal combustion engines - truck, jeep and sports car. The final results after denoising process are evaluated by four speech quality metrics. The trend of quality improvement depending on the SNR of the input noisy signal is examined. A possibility of using denoised speech signals for further speech features analysis is briefly discussed.