Tspeech-AI System Description to the 5th Deep Noise Suppression (DNS) Challenge

This report presents the development of Tencent AI Lab's personalized speech enhancement system for the ICASSP 2023 deep noise suppression (DNS) challenge, which includes the use of a modified band-split recurrent neural network (BSRNN) and a multi-frequency discriminator to improve the perceptual quality metrics such as personalized DNSMOS P.835 (PDNSMOS). The proposed system outperforms the baseline system by 0.047 and 0.036 final scores in the headset track and speaker phone track, respectively, and ranks the 3rd in both tracks.