Thông tin chung

  English

  Đề tài NC khoa học
  Bài báo, báo cáo khoa học
  Hướng dẫn Sau đại học
  Sách và giáo trình
  Các học phần và môn giảng dạy
  Giải thưởng khoa học, Phát minh, sáng chế
  Khen thưởng
  Thông tin khác

  Tài liệu tham khảo

  Hiệu chỉnh

 
Số người truy cập: 49,670,864

 Abnormal gait detection with one camera using Hidden Markov Model
Tác giả hoặc Nhóm tác giả: Nguyễn Trọng Nguyên, Huỳnh Hữu Hưng, Jean Meunier
Nơi đăng: The 11th IEEE-RIVF International Conference on Computing and Communication Technologies (RIVF 2015); Số: 1;Từ->đến trang: Accepted;Năm: 2015
Lĩnh vực: Công nghệ thông tin; Loại: Báo cáo; Thể loại: Quốc tế
TÓM TẮT
Studies on gait analysis play an important role in injury diagnosis, health care and in-home monitoring. We propose an approach for detecting abnormal gaits using a very simple setup of one camera instead of a complex system of calibrated cameras or depth sensors. The main idea is modeling the normal gait in each walk cycle via training patterns; the anomaly is then detected based on the likelihood of normality with a threshold which is determined automatically during the training process. Our processing consists of three main stages: preprocessing, feature extraction and recognition. The first stage involves two sub-stages: foreground extraction which determines the human silhouette using the combination of running Gaussian average and frame differencing techniques, and movement representation with motion history image. In the next stage, four characteristics are computed, and considered as a feature vector of each captured frame. Finally, each vector is converted into a codeword obtained with the k-means clustering technique, and hidden Markov model (HMM) is performed for modeling and recognizing. The experimental results show that our approach has good ability in distinguishing normal and abnormal gaits.
ABSTRACT
Studies on gait analysis play an important role in injury diagnosis, health care and in-home monitoring. We propose an approach for detecting abnormal gaits using a very simple setup of one camera instead of a complex system of calibrated cameras or depth sensors. The main idea is modeling the normal gait in each walk cycle via training patterns; the anomaly is then detected based on the likelihood of normality with a threshold which is determined automatically during the training process. Our processing consists of three main stages: preprocessing, feature extraction and recognition. The first stage involves two sub-stages: foreground extraction which determines the human silhouette using the combination of running Gaussian average and frame differencing techniques, and movement representation with motion history image. In the next stage, four characteristics are computed, and considered as a feature vector of each captured frame. Finally, each vector is converted into a codeword obtained with the k-means clustering technique, and hidden Markov model (HMM) is performed for modeling and recognizing. The experimental results show that our approach has good ability in distinguishing normal and abnormal gaits.
© Đại học Đà Nẵng
 
 
Địa chỉ: 41 Lê Duẩn Thành phố Đà Nẵng
Điện thoại: (84) 0236 3822 041 ; Email: dhdn@ac.udn.vn