Speech segregation based-on binaural cue: interaural time difference (itd) and interaural level difference (ild)

8th International Conference on Physics and its Applications

Farid, Mifta Nur^1 ; Arifianto, Dhany^1

Dept. of Engineering Physics, Faculty of Industrial Technology, Institut Teknologi Sepuluh Nopember, Kampus ITS Sukolilo, Surabaya

60111, Indonesia^1

关键词: Cocktail party effects;    Cross correlation methods;    Human auditory system;    Interaural level differences;    Interaural time differences;    Optimal performance;    Sound source separation;    Using probabilities;   

Others  :  https://iopscience.iop.org/article/10.1088/1742-6596/776/1/012067/pdf
DOI  :  10.1088/1742-6596/776/1/012067

来源: IOP

【 摘 要 】

A person who is suffering from hearing loss can be helped by using hearing aids and the most optimal performance of hearing aids are binaural hearing aids because it has similarities to human auditory system. In a conversation at a cocktail party, a person can focus on a single conversation even though the background sound and other people conversation is quite loud. This phenomenon is known as the cocktail party effect. In an early study, has been explained that binaural hearing have an important contribution to the cocktail party effect. So in this study, will be performed separation on the input binaural sound with 2 microphone sensors of two sound sources based on both the binaural cue, interaural time difference (ITD) and interaural level difference (ILD) using binary mask. To estimate value of ITD, is used cross-correlation method which the value of ITD represented as time delay of peak shifting at time-frequency unit. Binary mask is estimated based on pattern of ITD and ILD to relative strength of target that computed statistically using probability density estimation. Results of sound source separation performing well with the value of speech intelligibility using the percent correct word by 86% and 3 dB by SNR.