In　this　study,　we　create　a　mapping　model　between　surface　electromyography　(sEMG),　mechanomyography　(MMG),　and　joint　motion　angles　using　a　temporal　convolutional　network　(TCN)　to　estimate　elbow　joint　movement.　The　proposed　method　improves　the　accuracy　of　human　motion　estimation　based　on　the　complementary　nature　of　sEMG　and　MMG　signals　and　the　superior　accuracy　and　efficiency　of　the　TCN　in　sequence　modeling　and　capturing　time-dependent　information.　During　the　experimenter＇s　elbow　flexion　and　extension　movements,　we　collect　four-channel　sEMG　and　two-channel　MMG　signals.　The　four　eigenvalue　sequences　of　root-mean-square,　absolute　mean,　wavelength,　and　number　of　over-zero　points　are　extracted　as　the　training　set　of　the　TCN　model　after　alignment　and　filtering　processes,　and　the　prediction　results　are　compared　with　the　back-propagation　neural　network　(BPNN).　The　experimental　results　validate　the　effectiveness　of　the　proposed　method　and　its　advantages　in　comparison　with　the　related　results.　©　2024　IEEE.