Low-Dimensional Bottleneck Features for On-Device Continuous Speech Recognition
This work addresses battery usage reduction for on-device speech recognition systems, but it is incremental as it builds on existing bottleneck feature methods.
The paper tackled the problem of limited memory on low-power DSPs for continuous speech recognition by developing efficient bottleneck feature extractors that reduce feature size with minimal accuracy loss, achieving a 10x reduction without word error rate increase and a 64x reduction with only a 5.8% relative increase.
Low power digital signal processors (DSPs) typically have a very limited amount of memory in which to cache data. In this paper we develop efficient bottleneck feature (BNF) extractors that can be run on a DSP, and retrain a baseline large-vocabulary continuous speech recognition (LVCSR) system to use these BNFs with only a minimal loss of accuracy. The small BNFs allow the DSP chip to cache more audio features while the main application processor is suspended, thereby reducing the overall battery usage. Our presented system is able to reduce the footprint of standard, fixed point DSP spectral features by a factor of 10 without any loss in word error rate (WER) and by a factor of 64 with only a 5.8% relative increase in WER.