Generative AI Empowers Minimalist Wearable Personalized Human-Machine Interface
The seamless integration of electronics with the human body is pivotal for next-generation human-machine interfaces (HMI) and personalized healthcare. Traditional high-density sensor arrays, while capable of capturing complex biomechanical data, impose significant power and comfort penalties. This study introduces the Generative EMG Network (GenENet), a framework that synergizes generative artificial intelligence with soft bioelectronics to reduce hardware complexity. By leveraging a 32-channel stretchable sensor array as a 'teacher' dataset, GenENet employs a masked autoencoder architecture to learn spatiotemporal correlations within high-density electromyography (EMG) data. The trained model enables a simplified 6-channel wearable band to replicate the performance of the full 32-channel array. The sensor device utilizes a polydimethylsiloxane (PDMS) substrate, liquid metal (EGaln) interconnects, and a conductive PEDOT:PSS hydrogel interface, achieving low skin-contact impedance and high signal-to-noise ratios under mechanical strain. This approach addresses the bottleneck of data throughput and power consumption in wearable HMIs, offering a path toward minimalist, personalized devices for applications such as sign language decoding and gait analysis. The findings underscore the potential of generative AI to transform wearable bioelectronics by shifting computational burden from hardware to software.