The Research On Objective Strategies Of Speech Intelligibility

Under the background of information age, the voice is the most direct form of information transmission and exchange in people’s daily life. However, most of the speech is often accompanied by noise and make the people decline of comfortable degree in real life. It is difficult to understand the voice of the people who are hearing impaired with different background noise. Therefore, how to separate the noise, improve the intelligibility and evaluate the intelligibility from the signal has become an important problem.At present, the technology of ideal binary mask (Ideal Binary Mask, IBM) is proposed to provide hope for the evaluation of speech intelligibility under the noise background. Recent studies indicate that the importance of each time-frequency (T-F) unit to the intelligibility of the speech is related to the speech content. T-F units are categorized into two classes, speech-present T-F units and speech-absent T-F units. The experiment results in this paper show that the importance of speech intelligibility and the loudness of the target speech related in each voice of T-F unit.In 2008, Li and Loizou studies that, false alarm errors are shown to be more harmful to speech intelligibility than miss errors when the mixture signal-to-noise ratio (SNR) is below 5dB. In this experiment, the input mixed SNR is expanded, and the effects of two kinds of masking errors on the intelligibility of speech are studied under different input SNR. False alarm errors are shown to be more harmful to speech intelligibility than miss errors when the mixture signal-to-noise ratio (SNR) is below OdB. Recent studies on binary masking techniques provide the assumption that each time-frequency units contribute an equal amount to the overall intelligibility of speech and unchanged the structure of IBM. Considering that, we put forward a method to control the false alarm errors, miss errors based on weight, combining with the change of IBM structure in this paper, which is intended to put forward a better evaluation of the index of speech intelligibility.