A Comparative Study Of Sensor Activity Recognition Based On Shallow Learning And Deep Learning

With the development of artificial intelligence,the research of human activity recognition has been recognized.There are many challenges in the development of activity identification,such as the design,implementation and performance evaluation of activity recognition system.The development of mobile terminal can be described as advancing rapidly,which brings great convenience to people's lives.Therefore,the goal of this paper is to study the human activity recognition system based on wearable sensor data.The traditional machine learning algorithm has achieved good results in the field of human activity recognition,but a large number of manual extraction features make the recognition effect is not so smart,and the experimental process is cumbersome.With the development of machine learning,there is a branch of machine learning and deep learning the arrival of the liberation of the original hand can automatically extract features from the original data learning characteristics,and realizes the recognition of end-to-end mode.Firstly,this paper introduces the principle of machine learning,and introduces the basic process of human activity recognition.The experiments on human activity recognition data set collected by sensors are carried out.To introduce a wearable sensor data acquisition of human activities,the time signal of time series segmentation of signal segmentation window size,the original data were manually extracted different feature types,the effects of different classifiers on the performance of the recognition system.The experimental results show that it is very important to select the correct classifier to extract the appropriate features.Secondly,this paper introduces the introduction of deep learning,mainly studies the deep learning of the convolution neural network for human activity recognition.The principle of convolutional neural network and its components are introduced.Experimental verification of the application of convolutional neural network in the field of human activity recognition.The experimental results show that the convolutional neural network can automatically learn the features from the original data,and achieve the end to end pattern recognition.Finally,this paper introduces a gait recognition system based on convolution neural network for patients with Parkinson's disease.According to the characteristics of gait freezing in Parkinson's disease,a suitable convolution neural network is constructed.The experimental results show that the convolution neural network is applied to the gait recognition of patients with Parkinson's disease.