論文ID: 22.20250407
There is a contradiction between the number of communication channels and the number of sensors for wearable sensing systems. Increasing the number of sensors improves monitoring accuracy and function but makes communication difficult because of the increased channel necessity. In this paper, we developed a multi-sensor wearable system with only one single transmission channel and applied it to running gesture monitoring. By combining the k-nearest neighbor (kNN) machine learning method for signal analysis, we achieved classification of different running gestures. The implementation of a single signal transmission channel is based on hardware-level amplitude modulation, which is realized by designing the flexible piezoelectric Polyvinylidene fluoride (PVDF) sensors into various sizes. The different sizes enable amplitude modulation of the sensed signals. The modulated signals from different sensors are merged into a single-channel signal and transmitted to a personal computer via a wireless transmission circuit powered by a piezoelectric energy source inside the system. By utilizing the kNN algorithm, we successfully classified signals with different characteristics. Ultimately, two distinct running gestures were successfully detected and differentiated. This design presents an effective method to reduce signal transmission pathways and energy consumption, while also demonstrating that artificial intelligence algorithms can efficiently analyze data to extract useful information. This design method may provide broad inspiration for the development of wearable devices and holds significant promise for sport sensing.
