Women Safety Device Designed Using IoT and Machine Learning

Women safety is a very important issue due to rising crimes against women these days. Presently there is indeed no good solution to this problem. The existing applications and devices are not much effective as they need lot of human interaction to operate. These existing devices use to read the human temperature and heartbeat to generate alarm in case of emergency. When a person runs, every human may have different body temperature and heartbeat pattern and thus keeping a fixed threshold for finding out emergency situation and then generating alarm is not correct way and this is where the existing devices are failing to correctly generate alarm in case of emergency. In this paper the devices are customized to learn the individual pattern of temperature and heartbeat and then it finds out the threshold for generating alarm. Thus this paper deals to design a wearable women safety device that automatically reads and create patterns such as body temperature and pulse rate during running. If readings are higher than the normal readings then it will automatically call and message more than one person along with the location so that actions can be taken. We have used temperature and pulse sensors that will detect the activity of the woman and that data of sensors will be sent to cloud where machine learning algorithm (logistic regression) is applied to analyse the data generated. The data is first collected by sensors in non-danger conditions to train the algorithm, after that data is used for testing to gauge the accuracy and how close it is to our trained data. More is the accuracy more is the surety of danger and the emergency alarm will be there on emergency contacts. Thirdly, this paper deals with scenarios where there is no internet facility. To overcome the problem of internet we have used ZigBee mesh network, which helped the device to send the data to multiple hop distance.