Fall detection (human activity recognition) form accelerometer signals in x, y, z directions. Fall detection is an emerging field of scientific research in many domains like health care, human-machine interactions, human activity in industrial operations, and object motion activity in many industrial applications.
Present days, in elderly people without care takers at home is at high risk and mostly in stroke patients risk is more severe than the other people. To reduce such danger and risk, wearable devices are playing vital role in healthcare to notify immediately without any delays.
To address this issue, we developed a machine learning based fall detection to notify immediately not less than 10 seconds of time.
Fall detection (human activity recognition) is based on classifying accelerometer sensor data. Accelerometers are widely used in human activity recognition due to their compact nature, low power usage, and capacity to provide data directly related to the human motions/activities.
Data acquisition from accelerometer sensor data related to five human activities such as sitting, walking, fall while walk, fall from a chair, and fall while standing. Data preprocessing, feature engineering, model learning,and model performance with 10-fold cross validation method, and final model selection based on model accuracy forfall detection.
Fall detection system (human activity) from accelerometer Sensor data using supervised machine learning classification algorithm K-Nearest Neighbor (KNN):
• Fall detection from accelerometer sensor data using K-Nearest Neighbor (KNN) classifier.
• This application requires a training data from each human activity (fall or no fall).
• Application test with testing data from unknown human activity (fall or no fall).
The Video shows the how the activity of a person is recognized whether he is sitting, standing or jogging. The LED color changes based on the activities.The QueSSence Platform which is called the Intelligent Platform will be connected through Wi-Fi and Bluetooth will send the output at the edge device thus reducing the time taken to notify the concerned person regarding the activity.