Acoustic Echo Cancellation And Blind Speech Signal Separation Along With The DSP Implementation

The existence of the acoustic echo would badly affect communication quality and system stability. Therefore, acoustic echo cancellation (AEC) is necessary for the communication systems including Hand-free equipments and teleconferencing systems. Blind signal separation (BSS) is applied to a large number of areas such as array signal processing, teleconferencing system and speech enhancement.First the principle of stereophonic acoustic echo cancellation (SAEC) and the nonuniqueness problem caused by the correlation of the two input signals are introduced. Adaptive algorithms such as least mean square algorithm, normalized least mean square algorithm, extended least mean square algorithm, block frequency least mean square algorithm and fast recursive least square algorithm are applied to a real-time SAEC system. The misalignment and track performance of SAEC adaptive algorithms are investigated when the impulse responses of transmission room are changed abruptly in learning process.A simplified BSS model is deduced from the general model. An improved frequency domain blind separation algorithm for convolved speech signals is introduced. This frequency algorithm utilizes the second order statistic property and the nonstationarity of speech signals. It can extract the independent sources from the sensor signals by updating the coefficients of unmixture filters until the BSS outputs are uncorrelated with each other. A method using nonstationarity of speech signals to...