TY - JOUR
AU1 - Mohanaprasad, K.
AU2 - Soni Ishwarya, V.
AU3 - Ponraj, Abraham Sudharson
AU4 - Kar, Asutosh
AB - This paper proposed a new polynomial eigenvalue decomposition (PEVD)-based adaptive independent component analysis (ICA) for acoustic echo cancellation (AEC) during double-talk situation. The cancellation of acoustic echo using an adaptive filter fails during the double-talk situation. In this proposed method, the acoustic echo signal is removed from the near-end speech signal during the double-talk case without an explicit double-talk detector by using adaptive ICA techniques. The PEVD is a pre-processing phase before the implementation of adaptive ICA techniques. The PEVD is used to decorrelate the echo signal from the near-end signal, and it is also more effective with convolutive mixtures. The proposed method incorporates the PEVD pre-processing phase for AEC which aims at increasing the performance and it also compares the performance of echo cancellation with other singular value decomposition- and eigenvalue decomposition-based ICA techniques. Furthermore, the suggested technique is similarly compared to the traditional time-domain, frequency-domain adaptive filters, frequency-domain Kalman filter, and PEVD-based adaptive Kalman filter. Our simulation result shows that when compared with other pre-processing methods, the proposed PEVD-based ICA yields higher echo cancellation and less computation time.
TI - PEVD-Based Adaptive ICA for Acoustic Echo Cancellation During Double-Talk Situation
JF - "Circuits, Systems, and Signal Processing"
DO - 10.1007/s00034-022-01956-1
DA - 2022-06-01
UR - https://www.deepdyve.com/lp/springer-journals/pevd-based-adaptive-ica-for-acoustic-echo-cancellation-during-double-QsTczJzg0T
SP - 3570
EP - 3591
VL - 41
IS - 6
DP - DeepDyve
ER -