.png){kind=icon}
返回
MEMORY-EFFICIENT SPEECH RECOGNITION ON SMART DEVICES
.png){kind=icon}
OA
AI
DOI：10.1109/ICASSP39728.2021.9414502.png)
.png){kind=icon}
.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Recurrent transducer models have emerged as a promising solution for speech recognition on the current and next generation smart devices. The transducer models provide competitive accuracy within a reasonable memory footprint alleviating the memory capacity constraints in these devices. However, these models access parameters from off-chip memory for every input time step which adversely effects device battery life and limits their usability on low-power devices. We address transducer model's memory access concerns by optimizing their model architecture and designing novel recurrent cell designs. We demonstrate that i) model's energy cost is dominated by accessing model weights from off-chip memory, ii) transducer model architecture is pivotal in determining the number of accesses to off-chip memory and just model size is not a good proxy, iii) our transducer model optimizations and novel recurrent cell reduces off-chip memory accesses by 4.5x and model size by 2x with minimal accuracy impact.
Keyword:
RNN-T
ASR
Recurrent Transducer
Automatic Speech Recognition
On-device Inference
AI总结
对已上传原文的论文进行重点信息的提取，主要内容包括：简要概述、研究摘要、背景介绍、关键亮点、图文解析、展望与总结。
期刊
机构
暂无机构信息
引用论文
分享收藏Evolutionary Conservation of Orthoretroviral Long Terminal Repeats (LTRs) and ab initio Detection of Single LTRs in Genomic Data.png)
PLoS ONEOAAI
PLoS ONEIF0
OAAI分享收藏Efficient Processing of Deep Neural Networks: A Tutorial and Survey深度神经网络的高效处理: 教程和调查PROCEEDINGS OF THE IEEE
OAAI
PROCEEDINGS OF THE IEEEIF25.9
OAAI分享收藏 没有更多内容
没有更多内容