Name
Title | ||
|---|---|---|
Constructing ensembles of dissimilar acoustic models using hidden attributes of training data |
Abstract | ||
|---|---|---|
One of the objectives in acoustic modeling is to realize robust statistical models against the wide variety of acoustic conditions that are present in real world environments. As large amounts of training data become available, modeling subsets of the data with similar acoustic qualities can be done accurately and multiple acoustic models are jointly used as a form of system combination or model selection. In this paper, we propose a method to partition the training data for constructing ensembles of acoustic models using metadata attributes such as SNR, speaking rate, and duration via a binary tree. The metadata attribute used at each binary split in the decision tree is obtained using a metric proposed in this paper that is cosine-similarity based. The resulting multiple models are combined using voting techniques such as n-best ROVER. The proposed method improved the recognition accuracy by up to 4% relative over the state-of-the-art system on a large vocabulary continuous speech recognition voice search task. |
Year | DOI | Venue |
|---|---|---|
2012 | 10.1109/ICASSP.2012.6288830 | Acoustics, Speech and Signal Processing |
Keywords | Field | DocType |
decision trees,speech recognition,vocabulary,acoustic modeling,acoustic qualities,binary tree,continuous speech recognition,cosine-similarity,decision tree,dissimilar acoustic models,ensemble construction,hidden attributes,large vocabulary,n-best ROVER,recognition accuracy,robust statistical models,speaking rate,training data,voice search task,voting techniques,Automatic speech recognition,large corpora,multiple acoustic modeling,system combination | Decision tree,Metadata,Pattern recognition,Computer science,Binary tree,Model selection,Artificial intelligence,Statistical model,Hidden Markov model,Voice search,Binary number | Conference |
ISSN | ISBN | Citations |
1520-6149 E-ISBN : 978-1-4673-0044-5 | 978-1-4673-0044-5 | 0 |
PageRank | References | Authors |
0.34 | 9 | 4 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Takeshi Fukuda | 1 | 517 | 333.26 |
| Tachibana, R. | 2 | 20 | 7.01 |
| Chaudhari, U. | 3 | 13 | 1.70 |
| Bhuvana Ramabhadran | 4 | 1779 | 153.83 |