A tensorflow implementation of the Attention-based End-to-End Models for Small-Footprint Keyword Spotting
see requirements.txt
- feature: generate fbank or other feature, save to tfrecord format (todo)
- training: run train.sh
- inference: run infer.sh
- evaluate: (todo)
Modify the original model to streaming model which can support random length inputs.
- model: gru & lstm
- loss: focal loss & cross entropy loss & kl loss
Changhao Shan, Junbo Zhang, Yujun Wang, Lei Xie Attention-based End-to-End Models for Small-Footprint Keyword Spotting.