shivoh/deepreinforcementlearning-ddpg-for-roboticscontrol

This is an implementation of Deep Reinforcement Learning algorithm named Deep determinsitic Policy Gradient (DDPG) to train a 4 DOF robotic Arm to reach the moving target. The action space is continous and the learned agent outputs the torque for the robot to move to the specific target location.

Innovative Matched Filter creation for ECG heart rate detection

Motivation

The heart rate analysis from an ECG is crucial in assessing the health condition of an individual. But the measured ECG tend to have noises such as the 50Hz interference and the DC. These noises make the detection of peaks inaccurate which in turn make the detection of heartrate inaccurate. Thus, removing the noises from the ECG is pertinent for the proper analysis of the ECG.

The finite impulse response (FIR) filter is created to remove the 50Hz interference as well as the DC from the signal. This filtered signal is given to the matched filter which uses a mathematical function that has a similar shape of a single ECG wave to detect the R-peaks in the filtered ECG signal. The output from the matched filter is then used to find the momentary heart rate over time. We used two different functions similar to a single ECG signal for the matched filter and compared the outcome of both and selected the best for the calculation of momentary heart rate.