KeywordSpotting
目录
Alvarez, Raziel, and Hyun-Jin Park. “End-to-end streaming keyword spotting.” ICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2019.
- keywords: keyword
Paper: keywrod spotting
Summary
- an efficient memorized neural network topology that aims at
making better use of the parameters and associated computations in the DNN
by holding a memory of previous activations distributed over the depth of DNN. - a method to train the DNN, to produce the keyword spotting score.
- outperform in terms of quality of detection as well as size and computation.
Functions
- Efficient memorized neural network topology
- the memory keeps each inference’s state isolated from subsequent runs, just pushing new entries and popping old ones based on the memory size T configured for the layer.
- by stacking the SVDF layers to extending the receptive field of the network.
- Method to train end-to-end neural network
- label generation
- input sequence pairs$<X_t,c>$, $x_t$ is the 1D tensor corresponding to log-mel filter-band energies produced by front end handle;
- label generation