Name
Abstract | ||
|---|---|---|
With the advance in semiconductor memory and the availabil- ity of very large speech corpora (of hundreds to thousands of hours of speech), we would like to revisit the use of discrete hidden Markov model(DHMM) in automatic speech recogni- tion. To estimate the discrete density in a DHMM state, the acoustic space is divided into bins and one simply count the rel- ative amount of observations falling into each bin. With a very large speech corpus, we believe that the number of bins may be greatly increased to get a much higher density than before, and we will call the new models, the high-density discrete hid- den Markov model (HDDHMM). Our HDDHMM is different from traditional DHMM in two aspects: firstly, the codebook will have a size in thousands or even tens of thousands; sec- ondly, we propose a method based on scalar quantization in- dexing so that for a d-dimensional acoustic vector, the discrete codeword can be determined in O(d) time. During recogni- tion, the state probability is reduced to an O(1) table look-up. The new HDDHMM was tested on WSJ0 with 5K vocabulary. Compared with a baseline 4-stream continuous density HMM system which has a WER of 9.71%, a 4-stream HDDHMM sys- tem converted from the former achieves a WER of 11.60%, with no distance or Gaussian computation. |
Year | Venue | Keywords |
|---|---|---|
2005 | INTERSPEECH | indexation |
Field | DocType | Citations |
Speech corpus,Scalar multiplication,Pattern recognition,Computer science,Search engine indexing,Gaussian,Vector quantization,Artificial intelligence,Quantization (signal processing),Hidden Markov model,Codebook | Conference | 5 |
PageRank | References | Authors |
0.61 | 5 | 4 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Brian Mak | 1 | 12 | 2.81 |
| Jeff Siu-Kei Au-Yeung | 2 | 64 | 7.75 |
| Yiu-Pong Lai | 3 | 8 | 2.18 |
| Manhung Siu | 4 | 464 | 61.40 |