A Cascade Architecture for Keyword Spotting on Mobile Devices
arXiv:1712.03603
Abstract
We present a cascade architecture for keyword spotting with speaker verification on mobile devices. By pairing a small computational footprint with specialized digital signal processing (DSP) chips, we are able to achieve low power consumption while continuously listening for a keyword.
arXiv comment31st Conference on Neural Information Processing Systems (NIPS 2017), Long Beach, CA, USA
Cited by in corpus (8)
- TensorFlow Lite Micro: Embedded Machine Learning on TinyML Systems
- Multi-task Learning for Voice Trigger Detection
- Surprisal-Triggered Conditional Computation with Neural Networks
- Dynamic curriculum learning via data parameters for noise robust keyword spotting
- End-to-End Streaming Keyword Spotting
- Stacked 1D convolutional networks for end-to-end small footprint voice trigger detection
- On Front-end Gain Invariant Modeling for Wake Word Spotting
- Multi-task Learning with Cross Attention for Keyword Spotting