When estimating the heart rate (HR) of an exerciser by using a photoplethysmographic (PPG) sensor, the PPG output is contaminated with motion artifact (MA) induced by his motion, resulting in erroneous HR. To cancel the MA in the PPG output, we have proposed a technique based on adaptive filter algorithm using an MA sensor. On the one hand, we have so far fixed the tap length of the adaptive filter algorithm for the sake of its simple implementation, but on the other hand, we have noticed that the tap length dynamically changes according to the type of sensor wearer, the intensity of exercise and so on. Therefore, in this paper, we propose an MA canceling-PPG HR sensor system based on a serially configured adaptive filter algorithm with variable tap length. Experimental results involving 13 subjects reveal that the MA canceling technique based on the proposed serial configuration outperforms that with a conventional parallel configuration, achieving the minimum root mean square error of 9.97 beats per minute with much less computational complexity.