The speech enhancement is a challenging research field with numerous applications. In this thesis, we analyze the performance of Kalman and $H_∞$ filtering algorithms in speech enhancements. The theories of Kalman and $H_∞$ filtering methods are investigated. We also compare the performance of both filtering methods in various cases of signal-to-noise ratios (SNRs).
Simulation results show that the Kalman filtering gives better performance than the $H_∞$ filtering in the case of the speech corrupted by White Gaussian noise, especially in low SNR while the $H_∞$ filtering gives better performance than the Kalman filtering in the case of the speech corrupted by colored noise in general. However, in the $H_∞$ filtering, the parameter adaptation is critical to get the good performance.