Falls pose a significant risk to the elderly as they can result in fatalities. Especially for people who currently suffering from serious illnesses. To prevent and minimize this issue, in this research, we developed and built a system capable of fall detection which sends notifications via a Telegram bot to physicians and the patient’s family members for timely intervention in emergency situations, thus reducing undesirable risks. We employed the MediaPipe Pose that detects the position of a subject using two datasets of 860 videos with 30 fps that were recorded in-door condition from different perspectives. The system acquires the locations of 33 body landmark that are associated with the human body’s joints points and detects the presence of human movement via the detection of point changes in association with the human body landmark locations. We experimented with various time-series LSTM models to capture changes in human body landmark locations over continuous time. The experimental findings indicate that the proposed model achieves a fall detection accuracy of 98.2%.