ON ROBUST KALMAN FILTERING WITH FORGETTING FACTOR FOR SEQUENTIAL SPEECH ANALYSIS

We propose a robust Kalman filter with forgetting factor to estimate t he time-varying parameters of speech signals. The proposed robust Kalm an filter is based on a modified least-squares criterion with forgetti ng factor. The input signal is assumed to have a heavy-tailed non-Gaus sian nature with outliers due to spiky excitation. To alleviate the ef fects of outliers, this algorithm extends the concept of Huber's min-m ax approach, named M-estimation, to the Kalman filtering. The introduc tion of forgetting factor enables the time-varying speech parameters t o be estimated, giving more weight on the most recent portion of the d ata. Experimental results show that the proposed algorithm achieves mo re accurate estimation and provides improved tracking performance. (C) 1997 Elsevier Science B.V.