Loading...
Please wait, while we are loading the content...
Similar Documents
Hardware Speech Recognition in Low Cost , Low Power Devices
| Content Provider | Semantic Scholar |
|---|---|
| Author | Kim, Sukun Nedevschi, Sergiu Patra, Rabin K. |
| Abstract | We investigate the problem of providing speech recognition functionality, as an alternative to visual userinterfaces, in hardware devices facing important power, cost and storage constraints. Our goal is to enable these devices to perform continuous real-time hardware speech recognition for languages with small vocabularies, up to tens of words, and having limited grammatical constructions. Real-time continuous speech recognition is a computationally-intensive, but highly parallelizable task, with most of the computation taking place in the decoding stage of speech recognition. To this end, we propose a custom hardware implementation of the decoder part of a HMM-based speech recognition system, minimizing cost and power consumption while maintaining a good recognition accuracy. We present our preliminary findings, using a simulator to evaluate our design. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://www.cs.berkeley.edu/~binetude/course/cs252/speech.pdf |
| Alternate Webpage(s) | http://www.eecs.berkeley.edu/~binetude/course/cs252/speech.pdf |
| Alternate Webpage(s) | http://www.sukunkim.com/research/course/cs252/speech.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
