Bio-signal processing is a major topic in the field of robotics. Especially, prostheses have been studied for a long time to detect amputee's intentions from bio-signals because they do not have their limbs. However, controlling prostheses using bio-signals such as from the brain and surface electromyograms (sEMG), is complex and nonlinear because these signals are noisy and varied. It is not easy to determine the extent of motion such as a joint angle, and previous research used complex models or machine learning based on bio-signals. To estimate a joint angle easily and accurately, we propose a new bio-signal derived from the muscle bulge movements measured on the skin. We hypothesized a simple relationship between the muscle bulge movement longitudinally along a muscle and the corresponding joint angle, because the muscle contraction causes the change in the joint angle. We estimated a wrist joint angle as a function of the muscle bulge movement longitudinally along the extensor carpi radialis longus that is the agonist muscle of wrist extension. From the experimental results, the following three achievements were obtained. First, we extracted a linear relationship from the raw data with a high determination coefficient. Second, we showed that the estimation error using the proposed method was not much different from that of using sEMG in related work. Third, we established that the proposed method could estimate the joint angle robustly for the external loads on a limb.