Method and apparatus for blind signal recovery in noisy, reverberant environments

摘要

A maximum-kurtosis, distortionless response (MKDR) technique and an extension, the maximum-kurtosis, Wiener estimate (MKWE) technique, are provided. In one form, blind estimates of the speech source's channel response are made from the microphone data and MVDR is applied. The source direction is estimated by finding weights that maximize output kurtosis, or the fourth central statistical moment, in the frequency domain. The MKWE approach approximates the Wiener filter by using MKDR-output noise power estimates to compute a Wiener post-filter. These approaches can be extended to block-adaptive versions if the speech source is not quickly moving in space.

主权项

1. A method, comprising:
receiving a sound input with a plurality of sound input sensors, the sound input comprising a target signal from a source and noise from at least one interferer; transforming the sound input into a frequency domain form represented by a plurality of different frequency bins; determining a plurality of recovery-filter weight sets as a function of kurtosis, each recovery-filter weight set corresponding to one of the frequency bins; determining a plurality of steering vectors, each steering vector corresponding to one of the frequency bins and one of the sound input sensors; determining a plurality of beamformers according to the steering vectors and the recovery-filter weight sets, each beamformer corresponding to one of the frequency bins; and providing an output signal representative of the target signal as a function of the sound input and the beamformers, wherein the steering vector comprises:ek,j:=Em⁡[Xk⁡[m]⁡[Tk*⁡[m]]j][Em⁡[Xk⁡[m]⁡[Tk*⁡[m]]j]]j; wherein k is a frequency bin index, wherein m is a segment or frame index, wherein X is the sound input, wherein T is the frequency domain representation of the source, wherein Em is an expectation operator with respect to m, and wherein j is a sensor index.