HUMAN ACTIVITY RECOGNITION USING DEEP LEARNING

The latest improvement of machines showing shrewd traits includes severa techniques, together with pc hardware and software program structure improvement. Many special hardware devices, wearable sensors, system gaining knowledge of, and deep gaining knowledge of version implementations are being implemented in human interest popularity (HAR) programs in latest times. However, to broaden excessive accuracy class structures for interest popularity the use of ilow-price hardware era is of large importance. To gain this goal, the examine makes use of sensor statistics from  low-price sensors, a gyroscope and an accelerometer, together with the implementation of an Artificial Neural Network (ANN) primarily based totally on deep gaining knowledge of version for HAR. The implementation makes use of 5 MLP fashions as base-newbies and a neural community as a meta-learner. Evaluation is accomplished on  special datasets that include the statistics for diverse human activities, like ‘walking’, ‘sitting’, ‘lying’,’standing’,’stairup’ and ‘stairdown’, etc. The collected statistics is from the accelerometer and Gyroscope sensors of the user’s smartphone. In the proposed system, we can recognize the current activities of the human being and provide an email alert when the person does the restricted activities. The alert tone is provided only in the real-time implementation of this project. The project also provides reports based on the activity that the person did. The alert will be provided through the email when the restricted activity selected on the registration time.