A MATLAB Simulink implementation of speech masking based on the MPEG psychoacoustic model 1

The MPEG standard for digital compression for high-fidelity audio provides two implementations of the psychoacoustic model, psychoacoustic model 1 and psychoacoustic model 2. The ISO/1EC 13818-3:1994(E) document contains descriptions of these models. This model is of interest for the development of speech enhancement algorithms that concentrate their effect on the perceptible components in a speech signal. MATLAB/Simulink software has been used to construct a system for implementing speech masking. It consists of two main parts; the estimation of masking threshold, and the application of masking including reconstructing the speech. The system has been successfully tested with pure tone and speech signals, and undergone some quality assessment testing using human subjects. The system will be used as a tool for investigating cross-channel interactions for binaural speech enhancement.