MFFNet-to-detect-fatigue-sEMG_

##The purpose of the experiment The main content of this warehouse comes from a work in our laboratory: using multi-dimensional feature fusion network to analyze sEMG data and detect muscle fatigue.

##Dependent environment

• torch -->1.1.0
• numpy -->1.17
• tenSorboard -->1.7
• cuda -->10.0
• torchsummary -->1.5.1

File introduce

• tools.py --> Custom loss function。
• DataHelper.py --> load train data and test data: numpy->torch.tensor
• train.py --> model train script
• Note --> Other documents(include model file) will be announced after the paper is published.

Datasets

• Dataset1 --> is the data of our laboratory, if you need it, you can contact us by email:address. Of course, it will be provided after the paper is published.
• Dataset2 --> is provided by Michalis et al[1]. and can be obtained through the contact method provided in the paper.

Data preprocessing

Here we show the processing results of dataset 1

• Filter

• STFT

Result

Here we show some results of the experiment

Reference

[1] M. Papakostas, V. Kanal, M. Abujelala, K. Tsiakas, F. Makedon, Physical fatigue detection through EMG wearables and subjective user reports: a machine learning approach towards adaptive rehabilitation, in: Proceedings 450 of the 12th ACM International Conference on PErvasive Technologies Related to Assistive Environments, PETRA 2019, Island of Rhodes, Greece, June 5-7, 2019, pp. 475-481.