In high-level training, core strength programs need to measurably present trunk stability, muscle synergy, load adaptation, and long-term response. This paper presents a framework for computer-aided scheme design and tracking of core strength training for professional athletes. Twenty-four weeks of data including 38,640 training records, inertial measurements, surface EMG, plantar pressure, heart rate variability, video skeleton points, and coach-rated action labels were collected from 86 athletes in sprint, basketball, soccer, and combat sports. Based on SVR, BiLSTM and attention-enhanced TCN, the framework constructs an intelligent scheme generation module, a multi-modal execution monitoring module and a long-term effect prediction module. Experimental results show that the accuracy of action quality classification is 93.4%, the mean absolute error of core stability score is 4.8%, and the F1 value of fatigue risk warning is 0.89. Compared with the empirical scheme, the system reduces the manual review time by about 41.6%, and supports the data-driven adjustment of training load, action combination, recovery interval and cycle plan in the training micro-cycle, forming a continuous and traceable training management record.
