$ cd ~/projects sEMG Gesture Recognition Using LSTM Networks Person-Independent — For Human-Computer Interaction ──────────────────────────────────────────────────── A paper proposing an approach using LSTM Neural Networks to detect gestures in sEMG data. tags: Python, Research, AI ──────────────────────────────────────────────────── ## abstract Hand gestures feel natural to perform, which makes them well-suited to use as Human-Computer Interaction interfaces. But detecting them with high accuracy in real-time is a challenging task. This paper presents an approach based on the Long Short-Term Memory Neural Network architecture to evaluate Surface Electromyography signals and determine the gesture performed. ──────────────────────────────────────────────────── ## conclusion Surface Electromyography (sEMG) is a convenient way to gather EMG data for HCI interfaces. Hyperparameter tuning is used to refine the model. Transfer Learning can be used to translate a pre-trained model to a new person with very little additional data and training time. The resulting NN was able to detect gestures with high accuracy. ──────────────────────────────────────────────────── ## Dataset and experimental scope The experiments use signals from the Myo armband to classify rock, paper, and scissors gestures. The accompanying dataset contains recordings from 13 subjects, with eight-channel EMG samples, orientation data, detected poses where available, and recording metadata. The paper studies how a model trained on existing subjects can adapt to a new person with a small amount of additional data. Its results belong to that gesture set and recording procedure; the paper and slides contain the model, tuning, and evaluation details. Myo gesture dataset on GitHub https://github.com/michidk/myo-dataset ──────────────────────────────────────────────────── $ ls ./links paper.pdf /files/projects/semg-gesture-lstm/paper.pdf slides.pdf /files/projects/semg-gesture-lstm/slides.pdf uni wiki https://wiki.tum.de/display/infar/%5B19WS+-+GR%5D+Person-Independent+sEMG+Gesture+Recognition+Using+LSTM+Networks+for+Human-Computer+Interaction dataset https://github.com/michidk/myo-dataset ──────────────────────────────────────────────────── $ ls ~/pages > about curl -L lohr.dev > projects curl -L lohr.dev/projects > blog curl -L lohr.dev/blog > stats curl -L lohr.dev/stats > contact curl -L lohr.dev/contact ──────────────────────────────────────────────────── https://lohr.dev