Research on speech enhancement evaluation of driver in vehicle

The evaluation index of speech enhancement quality for drivers is the key to ensure the performance of speech enhancement algorithm, but the existing methods for speech enhancement quality evaluation cannot accurately reflect the subjectivity of human sound perception. To solve this problem, the applicability of speech intelligibility index to the evaluation of speech enhancement algorithm is analyzed, and experiments on a certain brand of vehicle are conducted. The driver's voice signal is collected by building a uniform linear microphone array in the vehicle, and then the beamforming algorithm is used to enhance the voice signals of different microphone combinations in the array. The speech enhancement effects of different microphone combinations at different driving speeds are obtained and evaluated by using signal-to-noise ratio (SNR) and speech intelligibility index (SII), and the results show that the speech intelligibility index is more suitable for evaluating the performance of the speech enhancement algorithms.