Using spectral subtraction for suppression of noise in speech signals withanalog integrated circuits

A new modification of the spectral subtraction algorithm is presented which enables operating entirely in the time domain and is thus suitable for rea lization in analog integrated circuits, The noise spectrum is obtained duri ng speechless intervals and stored for spectral subtraction when speech is present in the signal. The frequency range of interest of the speech signal is divided into narrow frequency bands by means of a bank of band-pass fil ters, For each frequency band the noise model is realized as an auxiliary s ignal multiplied by a particular weight. A subsystem is presented that prod uces an output signal whose power is equal to the difference between the in put signal power and the noise model power for each frequency channel, ther eby realizing the spectral subtraction. Circuits to achieve the described o peration are outlined. Finally, simulation results of the noise removal alg orithm are shown in the form of a spectrogram and the results showing impro vement in automatic speech recognition an given.