Name
Abstract | ||
|---|---|---|
We describe a novel technique of SPLICE (Stereo-based Piecewise Linear Compensation for Environments) for high performance robust speech recognition. It is an efficient noise reduction and channel distortion compensation technique that makes effective use of stereo training data. We present a new version of SPLICE using the minimum-mean-square-error decision, and describe an extension by training clusters of hidden Markov models (HMMs) with SPLICE processing. Comprehensive results using a Wall Street Journal large vocabulary recognition task and with a wide range of noise types demonstrate the superior performance of the SPLICE technique over that under noisy matched conditions (19% word error rate reduction). The new technique is also shown to consistently outperform the spectral-subtraction noise reduction technique, and is currently being integrated into the Microsoft MiPad, a new generation PDA prototype |
Year | DOI | Venue |
|---|---|---|
2001 | 10.1109/ICASSP.2001.940827 | ICASSP |
Keywords | Field | DocType |
wall street journal,splice,speech recognition,noise,channel distortion compensation,microsoft mipad,hmm,large vocabulary speech recognition,least mean squares methods,noise reduction,word error rate reduction,noisy matched conditions,stereo training data,pda prototype,minimum-meansquare-error decision,robust speech recognition,mmse decision,spectral -subtraction noise reduction,hidden markov models,stereo-based piecewise linear compensation,training data,hidden markov model,acoustic noise,piecewise linear,word error rate,minimum mean square error | Noise reduction,Pattern recognition,Computer science,splice,Word error rate,Communication channel,Speech recognition,Artificial intelligence,Hidden Markov model,Distortion,Vocabulary,Piecewise linear function | Conference |
Volume | ISSN | ISBN |
1 | 1520-6149 | 0-7803-7041-4 |
Citations | PageRank | References |
60 | 6.06 | 6 |
Authors | ||
5 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Deng, Li | 1 | 9691 | 728.14 |
| A. Acero | 2 | 4390 | 478.73 |
| Li Jiang | 3 | 60 | 6.40 |
| Jasha Droppo | 4 | 861 | 68.35 |
| Xuedong Huang | 5 | 1390 | 283.19 |