Mostra el registre d'ítem simple
An ultra low-power hardware accelerator for automatic speech recognition
dc.contributor.author | Yazdani Aminabadi, Reza |
dc.contributor.author | Segura Salvador, Albert |
dc.contributor.author | Arnau Montañés, José María |
dc.contributor.author | González Colás, Antonio María |
dc.contributor.other | Universitat Politècnica de Catalunya. Departament d'Arquitectura de Computadors |
dc.date.accessioned | 2017-06-01T06:40:38Z |
dc.date.available | 2017-06-01T06:40:38Z |
dc.date.issued | 2016 |
dc.identifier.citation | Yazdani, R., Segura, A., Arnau, J., González, A. An ultra low-power hardware accelerator for automatic speech recognition. A: Annual IEEE/ACM International Symposium on Microarchitecture. "Proceedings of the 49th IEEE/ACM Symposium on Microarchitecture". Taipei: IEEE Press, 2016, p. 580-591. |
dc.identifier.isbn | 978-1-5090-3509-0 |
dc.identifier.uri | http://hdl.handle.net/2117/105093 |
dc.description.abstract | Automatic Speech Recognition (ASR) is becoming increasingly ubiquitous, especially in the mobile segment. Fast and accurate ASR comes at a high energy cost which is not affordable for the tiny power budget of mobile devices. Hardware acceleration can reduce power consumption of ASR systems, while delivering high-performance. In this paper, we present an accelerator for large-vocabulary, speaker-independent, continuous speech recognition. It focuses on the Viterbi search algorithm, that represents the main bottleneck in an ASR system. The proposed design includes innovative techniques to improve the memory subsystem, since memory is identified as the main bottleneck for performance and power in the design of these accelerators. We propose a prefetching scheme tailored to the needs of an ASR system that hides main memory latency for a large fraction of the memory accesses with a negligible impact on area. In addition, we introduce a novel bandwidth saving technique that removes 20% of the off-chip memory accesses issued during the Viterbi search. The proposed design outperforms software implementations running on the CPU by orders of magnitude and achieves 1.7x speedup over a highly optimized CUDA implementation running on a high-end Geforce GTX 980 GPU, while reducing by two orders of magnitude (287x) the energy required to convert the speech into text. |
dc.format.extent | 12 p. |
dc.language.iso | eng |
dc.publisher | IEEE Press |
dc.subject | Àrees temàtiques de la UPC::Informàtica::Arquitectura de computadors |
dc.subject | Àrees temàtiques de la UPC::Enginyeria de la telecomunicació::Processament del senyal::Processament de la parla i del senyal acústic |
dc.subject.lcsh | Automatic speech recognition |
dc.subject.lcsh | Microprocessors |
dc.subject.lcsh | Parallel processing (Electronic computers) |
dc.subject.other | Storage management |
dc.subject.other | Microprocessor chips |
dc.subject.other | Parallel architectures |
dc.subject.other | Power aware computing |
dc.subject.other | Power consumption |
dc.subject.other | Search problems |
dc.subject.other | Speech recognition |
dc.title | An ultra low-power hardware accelerator for automatic speech recognition |
dc.type | Conference report |
dc.subject.lemac | Processament de la parla |
dc.subject.lemac | Microprocessadors |
dc.subject.lemac | Processament en paral·lel (Ordinadors) |
dc.contributor.group | Universitat Politècnica de Catalunya. ARCO - Microarquitectura i Compiladors |
dc.identifier.doi | 10.1109/MICRO.2016.7783750 |
dc.description.peerreviewed | Peer Reviewed |
dc.relation.publisherversion | http://ieeexplore.ieee.org/document/7783750/ |
dc.rights.access | Open Access |
local.identifier.drac | 19685485 |
dc.description.version | Postprint (author's final draft) |
local.citation.author | Yazdani, R.; Segura, A.; Arnau, J.; González, A. |
local.citation.contributor | Annual IEEE/ACM International Symposium on Microarchitecture |
local.citation.pubplace | Taipei |
local.citation.publicationName | Proceedings of the 49th IEEE/ACM Symposium on Microarchitecture |
local.citation.startingPage | 580 |
local.citation.endingPage | 591 |