overview

Health is a very important factor in human life. To maintain a healthy body, exercise and sports are one of the most effective ways to stay healthy. The technology of today's era plays a huge role in helping us track our health data, such as workout cycles, measure our current heart rate, electrocardiogram, sleep tracker and so on. The goal of this project is to build a deep learning system capable of predicting a person's heart rate using personal data from wearable devices and environment data.

Heart disease is the top leading cause of death in the United States. According to the CDC, approximately 697,000 people in the United States died from heart disease in 2020—that’s 1 in every 5 deaths. Therefore, it is important for everyone to do an annual health checkup, and monitor their heart health in the early stage, to prevent and detect irregularities of the heart. With the advanced technology of wearable devices, people can track their heart rate regularly while working out. Moreover, with the deep learning system that we are researching and studying, including a trained deep learning architecture, will help predict the heart rate more accurately.

For Milestone 2, we conducted using Supervised Learning, a subcategory of Machine Learning. It is defined by its use of labeled datasets to train algorithms to classify data or predict outcomes accurately. The deep learning algorithms we will be using are neural networks to forecast the heart rate. Please visit our documentation page for more details on how we perform the training sets and how they can be assisted in monitoring heart rate.

For the final milestone, we continue to utilize supervised learning to generate data for the new models. Also, test different architectures for the new model, each architecture should be tested five times. The new modeling program is designed to split the input data into three different portions, location data, environment data, and time, speed, and heart rate data, and processed through two 1DCNN blocks. As a result, 1DCNN has the best accuracy among all the models we've been working with so far.