Single Channel Speech Enhancement Using Complex Kalman Filter in Noisy Reverberant Environments

Yang Liu, Yanmin Shi, Xinglu Ma
2018 Proceedings of the 11th EAI International Conference on Mobile Multimedia Communications   unpublished
The previous research has proved that restoring the instantaneous amplitude and phase individually by linear prediction with Kalman filter on Gammatone filterbank could improve quality and intelligibility of noisy reverberant speech significantly. However, this method still suffers from the double prediction errors caused by the separated estimation for instantaneous amplitude and phase. This paper aims to study feasibility for restoring both instantaneous amplitude and phase simultaneously by
more » ... omplex Kalman filter to improve previous method. The signal to error ratio (SER), perceptual evaluation of speech quality (PESQ) and SNR loss were used as objective evaluation metrics. Results showed that the proposed method could effectively improve these objective metrics more than the previous method.
doi:10.4108/eai.21-6-2018.2276600 fatcat:6o2yrs7nfva5rgkesc5xbjor2y