Future Research

     is recommended for this algorithm: Although this algorithm is effective, it is not perfect, and there are many places where it could be improved. For example, lowpass filtering or the heart rate to a lesser degree as the algorithm progresses could speed up the rate that it locks on, as it would filter out the initial frequency transients that are usually glitches. In addition, I would like to try using combs that are gaussians in width to improve how well a close but not quite there frequency can take effect. For similar effects, it might be useful to try and use a weighted sum of the different filter banks for the coefficient instead of simply selecting from one.

In addition, because it is relatively simple for a person to know which parameters must be tweaked to fix problems, it should be possible to write a recursive algorithm that also changes parameters such as the mean to get the best signal depending on the SNR.