8th International Conference on Spoken Language Processing

Jeju Island, Korea
October 4-8, 2004

Integration of N-best Recognition Results Obtained by Multiple Noise Reduction Algorithms

Takeshi Yamada, Jiro Okada, Nobuhiko Kitawaki

University of Tsukuba, Japan

During the last decade, a number of noise reduction algorithms were proposed for realizing noise robust speech recognition. However, their effectiveness strongly depends on noise conditions. One way for solving this problem is to select an optimal algorithm every time before or after recognition process. This paper proposes a new method for integrating N-best recognition results obtained by multiple noise reduction algorithms. The proposed method selects the best recognition result by using a confidence measure based on a frame-normalized log likelihood score. To evaluate the performance of the proposed method, recognition experiments were performed on the AURORA-2J connected digit recognition task. These results confirmed that the proposed method is very effective in the high and middle SNR conditions.

Full Paper

Bibliographic reference.  Yamada, Takeshi / Okada, Jiro / Kitawaki, Nobuhiko (2004): "Integration of n-best recognition results obtained by multiple noise reduction algorithms", In INTERSPEECH-2004, 141-144.