This paper presents BoostSole; a smart insole based system for automatic human gait recognition. It consists of a smart instrumented insole connected to the cloud via the patient’s smartphone using low-power wireless communication. First, the design of BoostSole is introduced with discussions...
This paper presents BoostSole; a smart insole based system for automatic human gait recognition. It consists of a smart instrumented insole connected to the cloud via the patient’s smartphone using low-power wireless communication. First, the design of BoostSole is introduced with discussions of sensors choice, placement, calibration, and data communication. Next, an adaptive multi-boost classification algorithm is deployed to accurately identify different gait patterns. The algorithm is fast and lightweight and can be implemented in ordinary smartphones with a small footprint in terms of computational requirements, energy consumption, and communication usage. Raw and on-device classified data can be securely uploaded to a distant cloud server for continuous monitoring and analysis. Indeed, they can be visualized and exploited by doctors to identify/correct walking habits and assess the risks of chronic pain associated with an abnormal walk. The system has been evaluated on a dataset containing three gait patterns, namely: shuffle walk; toe walking; and normal gait. Obtained results are promising with more than 97% classification accuracy accompanied by low response time and computational demands.
This paper presents BoostSole; a smart insole based system for automatic human gait recognition. It consists of a smart instrumented insole connected to the cloud via the patient’s smartphone using low-power wireless communication. First, the design of BoostSole is introduced with discussions of sensors choice, placement, calibration, and data communication. Next, an adaptive multi-boost classification algorithm is deployed to accurately identify different gait patterns. The algorithm is fast and lightweight and can be implemented in ordinary smartphones with a small footprint in terms of computational requirements, energy consumption, and communication usage. Raw and on-device classified data can be securely uploaded to a distant cloud server for continuous monitoring and analysis. Indeed, they can be visualized and exploited by doctors to identify/correct walking habits and assess the risks of chronic pain associated with an abnormal walk. The system has been evaluated on a dataset containing three gait patterns, namely: shuffle walk; toe walking; and normal gait. Obtained results are promising with more than 97% classification accuracy accompanied by low response time and computational demands.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.