Speech recognition on embedded systems requires components of low memory footprint and low computational complexity. In this paper a POS-based (part of speech based) language modeling approach is presented which decreases the number of language model parameters combined with a method for reducing memory consumptions via quantization of language model penalties. For the application of short message dictation a language model with about 10,000 words of vocabulary is generated. Using the POS-based language modeling approach the number of parameters comprises 70,058 penalties. The memory consumptions for storing those penalties are reduced about 50% using the presented coding method. Experiments show that the POS-based language model is able to reduce the WER up to 65% for n-best isolated word recognition in comparison to the case without language model. Moreover the increase of WER caused by coding of the language model penalties is not significant.
[1]
Sergey Astrov,et al.
Memory space reduction for hidden Markov models in low-resource speech recognition systems
,
2002,
INTERSPEECH.
[2]
Petra Witschel,et al.
Constructing linguistic oriented language models for large vocabulary speech recognition
,
1993,
EUROSPEECH.
[3]
Bernt Andrassy,et al.
Large vocabulary speaker independent isolated word recognition for embedded systems
,
2003,
INTERSPEECH.
[4]
Slava M. Katz,et al.
Estimation of probabilities from sparse data for the language model component of a speech recognizer
,
1987,
IEEE Trans. Acoust. Speech Signal Process..
[5]
Harald Höge,et al.
Experiments in adaptation of language models for commercial applications
,
1997,
EUROSPEECH.