People with hearing impairment will have great difficulty when communicating, let alone in noisy environments. However, current speech enhancement technologies leave much to be desired. This is because current enhancement process concentrates on the improvement of signal to noise ratio (SNR), which does not translate to speech intelligibility improvement.
People with hearing impairment will have great difficulty when communicating, let alone in noisy environments. However, current speech enhancement technologies leave much to be desired. This is because current enhancement process concentrates on the improvement of signal to noise ratio (SNR), which does not translate to speech intelligibility improvement. Developments in physiological and psychoacoustic research show that modulation space is closely related to the human perceptual system as speech quality mainly resides in the spectral envelopes of the modulation frequencies. This encouraging finding indicates that modulation domain carries information relating to speech intelligibility. This project will capitalise on the modulation space to identify and study key speech features, which affect speech intelligibility index. The challenge lies in characterising the identified modulation features from a spectral-temporal modulation viewpoint and to understand how they can be optimised to improve the overall speech intelligibility.