Deep learning for automatic stereotypical motor movement detection using wearable sensors in autism spectrum disorders
Abstract
A new application of deep learning in automatic SMM detection on wearable sensors. Feature learning via CNN outperforms handcrafted features in SMM classification. Parameter pre-initialization is useful to transfer knowledge in longitudinal data. Including temporal dynamics of the signal using LSTM improves the detection rate. Using an ensemble of LSTM learners provides more accurate and stable SMM detector.
- Publication:
-
Signal Processing
- Pub Date:
- March 2018
- DOI:
- 10.1016/j.sigpro.2017.10.011
- Bibcode:
- 2018SigPr.144..180M
- Keywords:
-
- Convolutional neural networks;
- Long short-term memory;
- Transfer learning;
- Ensemble learning;
- Wearable sensors;
- Autism spectrum disorders