This code can be used as the supplemental material for the paper: "Residual Spiking Neural Network on a Programmable Neuromorphic Hardware for Speech Keyword Spotting". (Published on IEEE ICSICT, October, 2022).
C. Zou, X. Cui, S. Feng, G. Chen, X. Wang and Y. Wang, "Residual Spiking Neural Network on a Programmable Neuromorphic Hardware for Speech Keyword Spotting," 2022 IEEE 16th International Conference on Solid-State and Integrated CIrcuit Technology (ICSICT), 2022, pp. 1-3, doi: xxx.
- This supplemental material gives a reproduction function of training and testing experiments of the improved residual RNN (R-SNN) in our paper. Totally, three kinds of optional residual architectures are compared.
README.md
- this readme file.data
- the speech dataset folder.figs
- the visualized figure folder.tensorlayer
- the provided binary/ternary package, named tensorlayer.tools
- some available scripts.k0B2_asr_shortcut_group4_noplace_A.py
- the training script for the traditional residual architecture i.e.A
in our paper.k0B2_asr_shortcut_group4_noplace_B.py
- the training script for the traditional residual architecture i.e.B
in our paper.k0B2_asr_shortcut_group4_noplace_C.py
- the training script for the improve residual architecture in our paper.Spiking_asr_shortcut.py
- the residual SNN inference script.spiking_ulils.py
- the tool script for various spiking operators.
- Python-3.6, librosa-0.4
- Tensorflow 1.2 for cpu or gpu
- CPU or GPU server
- Please note you have installed the package Tensorflow=1.2.x, then directly run with:
$ python k0B2_asr_shortcut_group4_noplace_C.py.py --k 0 --B 2 --learning_rate 0.01 --resume False --mode 'training'
for the improve residual architecture training, or
$ python Spiking_asr_shortcut.py --k 0 --B 2 --noise_ratio 0 --learning_rate 0.01 --resume True --mode 'testing'
for the improve residual architecture testing.
- Please refer to our paper for more information.
- There might be a little difference of results for multiple training repetitions, because of the randomization.
- Please feel free to reach out here or email: 1801111301@pku.edu.cn if you have any questions or difficulties. I'm happy to help guide you.