A customer designed pants that measure ECG (electro-cardiogram) during exercise. This is challenging, because the ECG signal is smaller away from the heart, and the motion of exercise interferes with picking up the tiny microvolt level signal. A further difficulty is that dry electrodes are far more convenient than wet electrodes, but it is more challenging to pick up a good signal with dry electrodes. The customer was not able to get satisfactory results. Voler was selected to create an algorithm that could recover the signal even with the movement distortion and the very low-level input signal.
In order to solve the problem, data was collected across several users doing different types of exercises and movements. All of this data was captured and analyzed in-depth. Tools were developed to allow characterization of the raw signal. Then, through analysis of the data, the algorithm was designed. Care had to be taken that the algorithm would work in a very low power environment, as battery life was key for this product. Given the constraints and the raw signal capture, an initial algorithm was designed and tested on captured data. Voler has a lot of experience with ECG signaling, from which an optimized algorithm could be designed with very high confidence. The algorithm was first designed in Matlab, which allows for rapid development and testing. After performance was attained, the algorithm was then converted into a C module that used the ARM CMSIS DSP libraries. The C module was also tested using Matlab to ensure that the algorithm and its implementation exactly matched. The C module was then integrated into the wearable and performed just as expected on the first attempt. Power consumption was very low to allow for long battery life. This project was completed on budget and on time.