Goal: A non-invasive method is proposed to identify in vivo the passive mechanical properties of deep soft tissues in the human leg. How: Force-displacement curves in response to a localized compression of the calf are measured with a custom made experimental setup. The material parameters of a finite element model are then calibrated against the experimental curves using a genetic algorithm. A thorough investigation of the efficacy of this method to identify such mechanical properties is conducted through a design of experiments analysis and mixed numerical - experimental validations. Results: It is the first time that a thorough analysis is conducted to really separate the contribution of deep and superficial tissues in the response to compression tests and this permits to estimate the parameters of deep soft tissues on four subjects independently of the response of their other tissues. Two strain energy density functions are compared. It is shown that a 2nd order reduced polynomial better describes the passive mechanical behavior of the deep soft tissues of the leg than the neo-Hookean model.