Enhancement of non-air conduct speech based on auditory masking

摘要

Besides air conduct speech, the non-air conduct speech may provide some exciting possibility of wide applications. This paper explores a new non-air conduct speech detecting method using millimeter wave radar. However, the combined noise which is introduced in the detecting system corrupted the non-air conduct speech greatly. Therefore, this study proposed an efficient way to enhance the corrupted speech by taking into account properties of the human auditory system in the speech enhancement process, it offers the possibility to adaptive adjust the tradeoff between noise reduction, residual noise and speech distortion. The results from both simulation and evaluation suggest that this algorithm is able to reduce the background noise efficiently and the residual noise is less structured while the distortion of MMW radar speech remains acceptable.