Human Activity Recognition Using Smartphone Sensors with Context Filtering

Abstract

In recent times, for different application of Ambient Intelligent services (e.g Smart home) ,Remote monitoring and Assisted healthcare , the use of smartphones for the recognition of human activities became a topic of very high interest. From different studies we had found that , simple activities like sitting, running, walking can be recognized easily but semi complex activity like ascending ,descending stairs, slow running or Jogging , fast running is often difficult to recognize accurately. For reducing error rate of recognizing these kind of activity ,we adopted a new method “context filtering” .We used heart rate data and barometric pressure sensor data as element of context filtering . Also we use DTW in a different way .In Normal DTW several templates are there for matching activities with corresponding templates but we use a steady state as templates for all activity and match every activity with this steady state and find a score then after applying KNN algorithm we find the optimum threshold value. After primarily classifying activity we used this context filtering method to correctly recognize activities. After completion of our study we have seen that accuracy level has increased precisely for similar kind of activities. We present here details of our activity detection system, its architecture and evaluation