Share on Facebook Tweet on Twitter Share on LinkedIn Share by email
Reverberated Speech Signal Separation Based on Regularized Subband Feedforward ICA and Instantaneous Direction of Arrival

Lae-Hoon Kim, Ivan Tashev, and Alex Acero

Abstract

In this paper, independent component analysis (ICA) in a sub-band domain has been extended into a feed-forward network. The feed-forward network maximizes mutual independence of separated current frames using information from both current and previous multi-channel frames of speech signals captured by a microphone array. To guide into a proper separation prevent-ing permutation and arbitrary scaling, we not only rely on the steered response for the first tap of the demixing filter but also penalize on the direction thus drastically increasing the mean squared error with the spatial filtered output. After conver-gence, by applying instantaneous direction of arrival (IDOA) based post-processing, we can additionally suppress the leakage of interference as well as the reverberated target signal. The signal to interference ratio (SIR) is improved more than 20 dBC for distances up to 2.7 m and angle differences down to 26°.

Details

Publication typeInproceedings
Published inInternational Conference on Acoustics, Speech and Signal Processing
PublisherIEEE
> Publications > Reverberated Speech Signal Separation Based on Regularized Subband Feedforward ICA and Instantaneous Direction of Arrival