During the operations, firefighters can be injured or killed because of the smoke and heat emission from the fire area, broken structure elements such as floors, walls, or boiling liquid ejection and gas explosion. Therefore, this paper aims to develop an efficient and portable system to monitor falls and high CO level through integrating a three degrees of freedom accelerometer and an MQ7 sensor to recorded acceleration and measured CO concentration with the embedded fall and high CO detection algorithms. The embedded fall detection algorithm can detect fall events with ultra-high accuracy without mistakenly identifying normal activities such as walking, standing, jogging, and jumping as fall events. The posture recognition and cascade posture recognition after three seconds are proposed in this paper to gain the accuracy of our proposed fall detection system. If a firefighter falls and is unable to stand up, the alert signal message will be sent to their commander outside through the GSM/GPRS module. The embedded high CO detection algorithm used to alert the dangerous CO level to recommend using self-contained breathing apparatuses (SCBA) and saving fresh air with acceptable CO level. We carefully investigated the proposed thresholds and window size before embedding them into the microcontroller. The sensitivity and accuracy achieved were around 96.5% and 93% respectively in our recorded data. Furthermore, the proposed fall detection algorithm also achieved higher geometric mean in comparison with Support Vector Machine classifier (SVM) and a nearest neighbor rule (NN) in the public datasets with the achieved around 99.44%, 98.41% and 95.76% respectively.
Read full abstract