The Identification Of Human Behavior And Its Impact On Health

With the rapid technological development in the field of sensor, the sensor-based smart wearable devices gradually come into public view. This smart wearable device can obtain real-time sensor signals and then recognize the current activities. And with the improved quality of life, people concern about healthy more than they ever in the past. Regular daily activities and physical health are closely related. Therefore, it is very important to recognize human activity in order to monitor the health status.The major goal of human activity recognition problem is to recognize the daily activities, such as walking, running, jumping etc. At prosent,there are some problems in the aspect of feature selection and model construction.To solve these problems, this paper proposed a new method of feature extraction and model construction, and further explore the relationship between the activities and the health. The contribution of this paper can be described as follows:1. In order to solve the subjectivity problems in feature selection of human activities, we proposed a feature fusion method based on principal component analysis. we explore the classfication effectiveness of priori feature(time domain and frequency domain feature) and non-priori feature. Then the priori and non-priori feature was fusioned by the PCA algorithm. This feature fusion method not only reduces the subjectivity and complexity of features, but also ensure the validity of data fusion and improve the recognition accuracy.2. Based on the sparsity of the training data in individuals and the poor generalization ability problem, we proposed a user-dependent model based on sparse representation. Based on cosine similarity between users, we update the over-complete dictionary which is compsosed of the similar user data. And the user-dependent model is built. The model can be maximized to reduce the lack of labeled data collected by new user and improve the generalization ability.3. Based on the data of exercise and sleep, we explore the relationship between the exercise and sleep. we explore the effectiveness of motion and sleep parameters in predicting healthy sleep. And then the mixed features of sleep and exercise were taken to the support vector machine to predict sleep quality. The results shows that the sleep health can be improved by human activity in some extent.