Chia-Yeh Hsieh

Novel Hierarchical Fall Detection Algorithm Using a Multiphase Fall Model

Falls are the primary cause of accidents for the elderly in the living environment. Reducing hazards in the living environment and performing exercises for training balance and muscles are the common strategies for fall prevention. However, falls cannot be avoided completely; fall detection provides an alarm that can decrease injuries or death caused by the lack of rescue. The automatic fall detection system has opportunities to provide real-time emergency alarms for improving the safety and quality of home healthcare services. Two common technical challenges are also tackled in order to provide a reliable fall detection algorithm, including variability and ambiguity. We propose a novel hierarchical fall detection algorithm involving threshold-based and knowledge-based approaches to detect a fall event. The threshold-based approach efficiently supports the detection and identification of fall events from continuous sensor data. A multiphase fall model is utilized, including free fall, impact, and rest phases for the knowledge-based approach, which identifies fall events and has the potential to deal with the aforementioned technical challenges of a fall detection system. Seven kinds of falls and seven types of daily activities arranged in an experiment are used to explore the performance of the proposed fall detection algorithm. The overall performances of the sensitivity, specificity, precision, and accuracy using a knowledge-based algorithm are 99.79%, 98.74%, 99.05% and 99.33%, respectively. The results show that the proposed novel hierarchical fall detection algorithm can cope with the variability and ambiguity of the technical challenges and fulfill the reliability, adaptability, and flexibility requirements of an automatic fall detection system with respect to the individual differences.

Wearable sensor-based activity recognition for housekeeping task

In order to improve healthcare services and support clinical professionals, it is important to develop the unobstructive and automatic ADLs monitoring system for healthcare applications. Currently, various works have been developed for the monitoring of daily activities, such as ambulation, kitchen task, food and fluid intake, dressing, and medication intake while only few works paid attention to the housekeeping task. Housekeeping activity is a complex task, generally important for the several clinical assessment tools. In this work, we design and develop a wearable sensor-based activity recognition system recognize housekeeping tasks and classify the activity level. The proposed system achieves 90.67% accuracy for housekeeping tasks recognition, and 94.35% accuracy for activity level classification, respectively. The results of the experiment demonstrate that the system is reliable and fulfills the requirements of the unobstructive, objective, and long-term monitoring system.

A machine learning approach to fall detection algorithm using wearable sensor

Falls are the primary cause of accidents for the elderly in living environment. Falls frequently cause fatal and non-fatal injuries that are associated with a large amount of medical costs. Reduction hazards in living environment and doing exercise for training balance and muscle are the common strategies for fall prevention. But falls cannot be avoided completely; fall detection provides the alarm in time that can decrease the injuries or death caused by no rescue. We propose machine learning-based fall detection algorithm using multi-SVM with linear, quadratic or polynomial kernel function, and k-NN classifier. Eight kinds of falling postures and seven types of daily activities arranged in the experiment are used to explore the performance of the machine learning-based fall detection algorithm. The emulated falls were performed on a soft mat by ten healthy young subjects wearing protectors. The k-nearest neighbor method with 0.1 second window size has the highest accuracy, which is 96.26%. The results show that the proposed machine learning fall detection algorithm can fulfill the requirements of adaptability and flexibility for the individual differences.

Smartphone-based Pedestrian localization algorithm using inertial and light sensors

To assist clinical profession for further diagnosis, patients are usually ordered to do additional examinations in the outpatient clinic process. However, patients are easily to get lost for a visit due to the complex environment and poor wayfinding in the hospital. In this work, the smartphone-based localization system is developed using inertial sensors and light sensor. The proposed positioning algorithm combines the motion model (Pedestrian Dead Reckoning and Footpath approaches) and the light model. The proposed positioning algorithm utilizes the step information, the map information, and the light information to improve the performance of Pedestrian Dead Reckoning and Footpath approaches.

Drinking gesture spotting and identification using single wrist-worn inertial sensor

Stroke is the primary cause of serious long-term disability in the world. A long period of a rehabilitation program is required for those patients with the function loss of upper limb motor. In order to track the progression of the rehabilitation, the approaches to assessment of upper limb performance is the important task to evaluate the effectiveness of therapies. However, the typical assessment approaches suffer some issues, such as subjective, time-consuming, human resource limitation. In this works, we develop the drinking activity monitoring system using wrist-worn inertial sensor for performance assessment of upper-limb movement. Such drinking activity monitoring system can support clinical profession to keep track of the progress and provide the adequate assistance for the patients. In the proposed drinking gesture monitoring system, the drinking gesture spotting model is proposed to observe the drinking gesture during daily living. The rule-based transition detection (RTD) model is proposed for identification of elementary motions including extension and flexion. The proposed drinking activity monitoring system have the 92% and 90% in accuracy for drinking gesture spotting and transition detection, respectively. Such results show that the proposed drinking activity monitoring using single wrist-worn sensor is reliable.

Human motion identification for rehabilitation exercise assessment of knee osteoarthritis

Osteoarthritis (OA) is one of the majority of chronic lower limb musculoskeletal conditions, affecting approximately 15% of the population. Rehabilitation exercise has been considered as a common and essential medical treatment for mild to moderate stages of knee OA. However, there are some issues and challenges should be tackled while OA patient performs rehabilitation exercise without supervision of therapist, such as improperly implement rehabilitation exercise and patient adherence. The objective of this study is to propose a machine learning-based human motion identification system to automatically classify rehabilitation types and the motion states. The overall accuracy for types recognition is 100% and for motion identification is 97.7%. The results show that the feasibility of the proposed human motion identification algorithm for home-based rehabilitation.

Hand gesture recognition for post-stroke rehabilitation using leap motion

In order to enhance and/or improve recovery after stroke, rehabilitation needs to start early and be monitored by continuous and recurrent long-term interventions in the clinic and home setting. The elderly is a high risk stroke group with advancing age, resulting in increasing demand of strengthened resource of hospitals and physiotherapist. The residential rehabilitation for stroke patients would effectively relieve shortages of medical resources. However, the residential rehabilitation for stroke patients faces with the lack of professional guidance, and physiotherapist cannot monitor the rehabilitation progress of stroke patients. These problems may lead to additional harm or deteriorate rehabilitation progress. In order to solve this problem, we develop a hand gesture recognition algorithm devoted to monitor the seven gestures for residential rehabilitation of the post-stroke patients. The gestures were performed by seventeen healthy young subjects. The results were assessed by k-fold cross validation method. The results show that the proposed hand gesture recognition algorithm using multi-class SVM and k-NN classifier achieve accuracy of 97.29% and 97.71%, respectively.