Please Read the Report for better understanding
Falls refer to sudden unintentional move from higher level to lower level without any control. Falls can happen with people of any age, gender at any time. But the main sufferers of fall event are the elderly people. Elder population is growing rapidly day by day around the world. Most of the elderly people remains unsupervised major part of the day. If a fall occur with any person, the person may be laying down for a long period of time without any help. Thus, a fall can be costly in terms of health, money and lives. Therefore, a fall detection system is needed to automatically detect fall events which will play an important role in health care system of elderly people. Recently many researchers have devoted themselves to develop systems and methods for automatic fall detection. Fall detection techniques can be divided in four groups based on the source of data. They are namely: life information-based methods, radar-based methods, wearable/mobile sensor based and vision-based methods. Wearable/mobile sensor-based methods can be used in both inside and outside of the room. Hence it is preferred most now a days. Because of the scarcity of publicly available sensor dataset, a dataset was constructed. It contains six activities namely Falling, Standing, Walking, Sitting, Sitting on chair in the back side and Laying. Accelerometer and gyroscope data were collected using mobile phone sensor. These data were used for the classification purposes. Wearable/mobile sensor-based fall detection system has the ability to detect fall both in outdoor and indoor. In this work, feature map was generated from the mobile sensor data. Mobile sensor data are mainly accelerometer and gyroscope data. Some features are extracted from the sensor data and feature map was generated using the features. Then this feature map was used to classify the fall action with other activities.
Methodology:
Diagram of CNN model used in this work:
Results:
Please Read the Report for better understanding