Forearm Radiomyography for Character and Writer Identification in Air Writing.

Area of Science:

• Human-Computer Interaction (HCI)
• Wearable Sensing Technology
• Biomedical Engineering

Background:

• Traditional HCI methods often involve hand-mounted devices or electrodes, which can be cumbersome or require wet contact.
• Near-field radio frequency (NFRF) sensing offers a potential alternative for unobtrusive human-computer interaction.
• Radiomyography (RMG) is an NFRF sensing modality capable of detecting muscle activity.

Purpose of the Study:

• To explore the application of radiomyography (RMG) for air writing recognition.
• To evaluate the efficacy of an NFRF sensing system for identifying handwritten characters and individual writers.
• To address key human-computer interaction (HCI) issues, including user acceptability and privacy concerns.

Main Methods:

• A forearm-worn multiple-input-multiple-output (MIMO) RMG system was developed using synchronized software-defined radios (SDRs) and custom antenna arrays.
• The system captured detailed muscle actuation during air writing motions via enhanced electromagnetic coupling.
• Signal processing and machine learning (ML) algorithms were benchmarked for classification accuracy and computational efficiency.

Main Results:

• The best performing ML algorithm achieved an average character classification accuracy of 92.35%.
• Transfer learning enabled 95% testing accuracy for new subjects with only 10% training data.
• The writer identification model achieved an average accuracy of 99.1% across longitudinal studies.

Conclusions:

• Forearm-worn RMG is a promising novel HCI alternative for accurate air writing recognition.
• This technology offers high performance without obstructing hand dexterity or requiring wet electrodes.
• RMG presents a viable solution for unobtrusive and effective human-computer interaction.