METHOD AND APPARATUS FOR DEREVERBERATION

In a method and an apparatus for dereverberation provided speech from speaker is received by a first and second channel microphones which are disposed at different locations, and is input to a first and a second channel reverberant speech input terminal. Input signal in each channel is processed by an inverse filter processor, and a dereverberation evaluation part evaluates dereverberation performance on the basis of an output signal from the inverse filter processor and the input signals of respective channels. Subsequently, filter coefficients in the inverse filter processor are determined and updated in accordance with the evaluation so that a result of evaluation is brought closer to an optimum. By repeating this process, an optimum dereverberation is always enabled in a manner following a variation in an in-room impulse response.