Abstract | ||
---|---|---|
A speaker adaptation scheme named maximum likelihood model interpolation (MLMI) is proposed. The basic idea of MLMI is to compute the speaker adapted (SA) model of a test speaker by a linear convex combination of a set of speaker dependent (SD) models. Given a set of training speakers, we first calculate the corresponding SD models for each training speaker as well as the speaker-independent (SI) models. Then, the mean vector of the SA model is computed as the weighted sum of the set of the SD mean vectors, while the covariance matrix is the same as that of the SI model. An algorithm to estimate the weight parameters is given which maximizes the likelihood of the SA model given the adaptation data. Experiments show that 3 adaptation sentences can give a significant performance improvement. As the number of SD models increases, further improvement can be obtained |
Year | DOI | Venue |
---|---|---|
1999 | 10.1109/ICASSP.1999.759777 | ICASSP |
Keywords | Field | DocType |
corresponding sd model,experiments,speech processing,linear convex combination,speech recognition,si model,adaptation sentences,interpolation,weighted sum,sa model,maximum likelihood estimation,adaptive signal processing,speaker dependent models,speaker-independent models,covariance matrices,sd models increase,sd mean vectors,maximum likelihood model interpolation,sd mean vector,training speakers,algorithm,test speaker,weight parameter estimation,speaker adaptation scheme,adaptation data,speaker adaptation,performance,covariance matrix,training speaker,mean vector,parameter estimation,natural languages,vectors,convex combination,hidden markov models,maximum likelihood,loudspeakers | Speech processing,Pattern recognition,Convex combination,Interpolation,Maximum likelihood,Artificial intelligence,Adaptive filter,Covariance matrix,Speaker adaptation,Mathematics,Performance improvement | Conference |
Volume | ISSN | ISBN |
2 | 1520-6149 | 0-7803-5041-3 |
Citations | PageRank | References |
5 | 0.93 | 4 |
Authors | ||
2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Zuoying Wang | 1 | 46 | 11.20 |
Feng Liu | 2 | 5 | 0.93 |