Speech enhancement technique in aircraft cockpit voice background

For flight accident investigation it is very important to efficiently extract the speech information against aircraft cockpit voice background.A speech enhancement method suited to the case of aircraft cockpit voice background was proposed.CHI PDF was chosen as the prior distribution for the statistical model of clean speech spectral magnitude;the minimum mean-square error short time spectral amplitude (MMSE-STSA) estimator was utilized to estimate the clean speech spectral magnitude;and a share function was used to give different emphasis on errors of small amplitudes relative to large amplitude in the error estimation function.In order to achieve the balance between noise reduction and speech distortion,the parameters in clean speech spectral estimation were chosen against the cockpit voice background by simulation experiments incorporating with the objective assessment of speech quality.Finally,in order to validate the effectiveness of proposed method,the comparison experiments between the classic methods and the proposed method were carried out.