Multi-channel time-frequency data fusion

Authors:

Abstract
This paper proposes an eficient mechanism for the fusion of two noisy speech signals obtained by an array of two microphones using single-tap time-frequency filters and by taking into account the correct time delay of arrival (TDOA) of the speech source. Speech signals ob- tained by the microphones are transformed into a set of two complex time-frequency (TF) images. By knowing the correct TDOA, and therefore the associated phase differ- ence between the signals at eacli frequency, it is possible to non-linearlyjlter both the real and the imaginary parts of the TF images. This will consist of a TF reward-punish filter that adjusts the amplitude of the TF blocks based upon the variation of their phase-difference with the ideal phase-difference dejned by the TDOA. Simulation results show that the proposed technique can achieve a Signal- to-Noise Ratio (SNR) improvement of 15dB when there is strong Gaussian noise present (-20dB initial SNR). When the original SNR is OdB, the simulated improvement is ap- proximately 8dB. It is also shown that although the pro- posed technique is a more general case of the adaptive beamfomier (where the adaptive beamformer has a spec@ reward-punish characteristic), other reward-punish char- acteristics that are proposed in this paper can ofren surpass the pelfonnance of the ideal adaptive beamformer.

This publication has 12 references indexed in Scilit: