Skip to main content

Gina Cody School researchers innovate in AI, improving hand gesture recognition technology

December 11, 2023
|

Arash Mohammadi Photo credit: David Ward

Sometimes less is more, even in the field of artificial intelligence and machine learning. Mansooreh Montazerin, recent master's student graduate under the supervision of Arash Mohammadi and Farnoosh Naderkhani at the Gina Cody School of Engineering and Computer Science, used a leaner machine learning approach to improve hand gesture recognition used in prosthetics, exoskeletons and mixed reality.

With colleagues from New York University (NYU) and University of Calgary, they relied on Vision Transformer architecture which uses an attention mechanism to identify patterns in data. It makes it suitable for analyzing muscle movements in the human arm using electrodes placed on the skin. 

Using this architecture, they created a framework that demonstrated good accuracy even with a smaller number of learnable parameters (65k) compared to traditional deep learning models which can reach millions of parameters.

Farnoosh Naderkhani Photo credit: David Ward

They describe their findings in Nature’s Scientific Reports, the 5th most cited journal in the world in 2022.

What sets their new framework apart is its learning efficiency. Unlike traditional models that require extensive pre-training, it can start learning immediately, much like a child picking up a language through observation and listening. This inherent learning capability makes the system both innovative and practical.

The efficacy of their new framework was tested as a first step with 20 able-bodied individuals, each performing 65 distinct hand gestures. The system's accuracy in interpreting these gestures was influenced by various factors, such as the number of sensors employed and the duration of observation. In some instances, the accuracy rate reached up to 92%, an achievement in the field of gesture recognition.

Future research might focus on using data from people with limb amputations and moving to real-time muscle movements analysis.

Mohammadi leads the Intelligent Signal & Information Processing (I-SIP) in the Concordia Institute for Information Systems Engineering.

Read “Transformer-based hand gesture recognition from instantaneous to fused neural decomposition of high-density EMG signals” in Nature’s Scientific Reports.

Learn more about the Concordia Institute for Information Systems Engineering.



Back to top

© Concordia University