A noise robust end point detection algorithm is proposed that could be used in real environment speech recognition. Inaccurate end point detection brings not only speech recognition performance reduction but also users’ tiredness. EPD algorithms based on energy level change or speech presence probability are vulnerable to high energy noises. After reducing much noise by auditory filter, one of human speech pronunciation characteristic, syllabic rate is used for checking if there is still speech component or not. The proposed algorithm shows much better performance in real environments like TV sound noise, café noise, etc.
(See document for exact affiliation information.)
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member Join the AES. If you need to check your member status, login to the Member Portal.
