Integrated sensing and communication (ISAC) is a promising technology to improve the band-utilization efficiency via spectrum sharing or hardware sharing between radar and communication systems. Since a common radio resource budget is shared by both functionalities, there exists a tradeoff between the sensing and communication performance. However, this tradeoff curve is currently unknown in ISAC systems with human motion recognition tasks based on deep learning. To fill this gap, this paper formulates and solves a multi-objective optimization problem which simultaneously maximizes the recognition accuracy and the communication data rate. The key ingredient of this new formulation is a nonlinear recognition accuracy model with respect to the wireless resources, where the model is derived from power function regression of the system performance of the deep spectrogram network. To avoid cost-expensive data collection procedures, a primitive-based autoregressive hybrid (PBAH) channel model is developed, which facilitates efficient training and testing dataset generation for human motion recognition in a virtual environment. Extensive results demonstrate that the proposed wireless recognition accuracy and PBAH channel models match the actual experimental data very well. Moreover, it is found that the accuracy-rate region consists of a communication saturation zone, a sensing saturation zone, and a communication-sensing adversarial zone, of which the third zone achieves the desirable balanced performance for ISAC systems.