In this study, we developed a low-cost, portable, and electrode-free intelligent device for the nondestructive evaluation of the muscle cross section of rehabilitation patients or elderly individuals. Currently, muscle quality is mostly explored by personal feelings. There is no convenient instrument that can quantify daily training progress on the local muscle level. Such a device can significantly encourage and motivate the elderly and rehabilitation patients to participate in a training program. Our sound injection resonant myophonogram can overcome the many challenges of existing muscle measurement devices. We actively injected a sound wave spectrum into the deep layers of the muscle to form a wavefront field. The reflected sound spectrum was then formed through the interaction of minute shear elasticity generated by the muscle tissue on the resonant point between the expansion and contraction of tissues. We employed an optimization algorithm on the generative adversarial network to learn the parameters of the muscle model and translated the responses into a performance index. To achieve such real-time predictive feedback, we implemented a sound excitation device and a cloud computing service to develop the algorithm with high performance at a low cost. Our device has been proven accurate and can perform measurements in real time. The device’s assessment achieved an accuracy of more than 90% in 0.5 s.