Scholarlink

Gait is a way of walking that is different for each individual. Gait can be used to recognize identity by using sensors or cameras. This paper presents individual recognition based on gait data obtained by gyroscope sensors using a machine learning algorithm, namely hidden markov model (HMM). This research is carried out by comparing the accuracy produced by models that use angular angle data only, magnitude, and a combination of angular angle and magnitude combined. The amount of segmentation data and hidden states used by the model also affects the accuracy of the model. The maximum accuracy achieved by the HMM is 97.5% with windowing 500 data and hidden states used 12 and using one sensor node and using only angular angle as a feature for training. When evaluating model performance, segmentation and the quantity of hidden states are interrelated. In the future, this system will be applied for gait biometrics authentication that can be implemented as a smart key model or algorithm.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Individual Identification Based on Gait Using Gyroscope Sensor and Hidden Markov Model Algorithm

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Abdillah Nur Isnaini,
  • Achmad Rizal,
  • Istiqomah,
  • Said Ziani

摘要

Gait is a way of walking that is different for each individual. Gait can be used to recognize identity by using sensors or cameras. This paper presents individual recognition based on gait data obtained by gyroscope sensors using a machine learning algorithm, namely hidden markov model (HMM). This research is carried out by comparing the accuracy produced by models that use angular angle data only, magnitude, and a combination of angular angle and magnitude combined. The amount of segmentation data and hidden states used by the model also affects the accuracy of the model. The maximum accuracy achieved by the HMM is 97.5% with windowing 500 data and hidden states used 12 and using one sensor node and using only angular angle as a feature for training. When evaluating model performance, segmentation and the quantity of hidden states are interrelated. In the future, this system will be applied for gait biometrics authentication that can be implemented as a smart key model or algorithm.