Name
Title | ||
|---|---|---|
Robust discriminative keyword spotting for emotionally colored spontaneous speech using bidirectional LSTM networks |
Abstract | ||
|---|---|---|
In this paper we propose a new technique for robust keyword spotting that uses bidirectional Long Short-Term Memory (BLSTM) recurrent neural nets to incorporate contextual information in speech decoding. Our approach overcomes the drawbacks of generative HMM modeling by applying a discriminative learning procedure that non-linearly maps speech features into an abstract vector space. By incorporating the outputs of a BLSTM network into the speech features, it is able to make use of past and future context for phoneme predictions. The robustness of the approach is evaluated on a keyword spotting task using the HUMAINE Sensitive Artificial Listener (SAL) database, which contains accented, spontaneous, and emotionally colored speech. The test is particularly stringent because the system is not trained on the SAL database, but only on the TIMIT corpus of read speech. We show that our method prevails over a discriminative keyword spotter without BLSTM-enhanced feature functions, which in turn has been proven to outperform HMM-based techniques. |
Year | DOI | Venue |
|---|---|---|
2009 | 10.1109/ICASSP.2009.4960492 | ICASSP |
Keywords | Field | DocType |
blstm network,discriminative keyword spotter,speech decoding,bidirectional lstm network,index terms— speech recognition,robust keyword,hmm-based technique,speech feature,robustness,blstm-enhanced feature function,recurrent neu- ral networks,spontaneous speech,non-linearly maps speech feature,humaine sensitive artificial listener,robust discriminative keyword,sal database,speech,speech recognition,long short term memory,discrimination learning,neural net,hidden markov models,speech coding,databases,logic gates,neural networks,hidden markov model,decoding,recurrent neural networks,indexing terms,vector space,computer science | Speech enhancement,TIMIT,Speech coding,Pattern recognition,Computer science,Recurrent neural network,Keyword spotting,Speech recognition,Artificial intelligence,Hidden Markov model,Artificial neural network,Discriminative model | Conference |
ISSN | Citations | PageRank |
1520-6149 | 29 | 1.34 |
References | Authors | |
9 | 6 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Martin Wollmer | 1 | 212 | 8.62 |
| Florian Eyben | 2 | 2854 | 141.87 |
| Joseph Keshet | 3 | 925 | 69.84 |
| Graves, Alex | 4 | 8572 | 405.10 |
| Björn Schuller | 5 | 6749 | 463.50 |
| Gerhard Rigoll | 6 | 2788 | 268.87 |