2020 HAR using Inertial, Physiological and Environment Sensors: a Comprehensive Survey

This survey reports the system review process with PRISMA methodology. It shows the more and more deep learning research on HAR compared to traditional machine learning method. The most frequent used sensor category is inertial sensor from 2015 to 2019, includes accelerometer, gyroscope and magnetometer. The other category is physiological sensor, includes Electroencephalogram(EEG), Electrocardiogram(ECG) and Electromyography(EMG). Another one is environmental sensor, includes temperature, humidity, pressure, brightness sensors and some sensors related to the people/objects present inside the environment.
1. There is the challenge issue of creating better generalization model and identifying a methodology superior to the others because of the lack of public heterogeneous datasets.
2. And at least 78% proposed results are not fully reproduciblr because of almost only 1/5 datasets are publicly available.