Hand Gesture Recognition Using Single Patchable Six-Axis Inertial Measurement Unit via Recurrent Neural Networks

Area of Science:

• Wearable technology
• Biomedical engineering
• Machine learning

Background:

• Wearable sensors are crucial for gesture control and healthcare.
• Miniaturized, patchable inertial measurement units (IMUs) offer improved comfort and reduced motion artifacts.
• Soft, adhesive substrates with serpentine interconnections enhance sensor adaptability to body deformations.

Purpose of the Study:

• To present a hand gesture recognition system utilizing a single, patchable six-axis IMU attached at the wrist.
• To develop and evaluate recurrent neural network (RNN) models for gesture recognition using IMU data.
• To demonstrate the potential of the proposed system for continuous motion monitoring in remote settings.

Main Methods:

• A novel patchable six-axis IMU was developed with IC-based components on a stretchable, adhesive substrate.
• The IMU incorporates wireless Bluetooth communication for continuous data transmission.
• Two RNN-based models were trained and validated using a public database to recognize three distinct hand gestures.

Main Results:

• The patchable IMU demonstrated effective data acquisition with minimized motion artifacts due to its soft form-factor.
• The RNN models achieved successful recognition of three hand gestures.
• Preliminary results indicate the system's potential for continuous motion monitoring.

Conclusions:

• The proposed patchable IMU system offers a comfortable and effective solution for hand gesture recognition.
• This technology has significant potential for applications in mobile health, human-computer interaction, and control gestures.
• Further development could enable widespread remote monitoring of human motion.