EMGSense: A Deep Learning-based EMG Sensing Technology for Wearables

AI researchers have published a new paper stating the development of a deep learning-based EMG Sensing Technology, called EMGSense. The developments were published in the paper titled “EMGSense: A Low-Effort Self-Supervised Domain Adaptation Framework for EMG Sensing”. The new human-machine interface (HMI) model fundamentally aims to solve the problem of cross-user EMG sensing, which is linked to the degradation of sensing devices due to biological heterogeneity.

Let’s understand the concept of EMG sensing technology and how a deep learning model like EMGSense works.

What is EMG Sensing Technology?

Electromyography or EMG sensing technology tracks and measures the electromyographic (EMG) signals that your muscles generate during locomotion. These sensors can detect faintest of muscle movements such as bending of elbows, rotating wrists, twiddling with the thumb, or lifting the arm. EMG sensing technology is used in the medical field to detect degenerative nervous or muscular disorders causing numbness, tingling, weakness and cramping. With the rise of smartphone-based connectivity sensors, the scope of EMG Sensing technology has grown beyond medical applications. Today, this technology is widely used to study biomechanics, to create customized prosthetic limbs, to build gesture-controlled immersive video games and robotic control systems.

There are two types of EMG Sensors;

• Surface
• Intramuscular

EMG sensing systems suffer from the problem of degradation and limited cross-user heterogeneity. Deep learning technology has a solution, as researchers from CityU- Hong Kong showed in their latest paper. The framework is called EMGSense– a high-performance self-trained model that delivers accurate output from wearables. Researchers developed advanced self-supervised deep neural network (DNN) model to measure EMG data from gesture recognition and activity recognition. EMGSense is much more accurate than other EMG sensing technologies, outperforming other models by up to 17.4%.