Ambulatory motion sensors can generate a mobile system for human kinematics measurement as an alternative to the more constrained optical motion capture systems. Flexible inertial sensors are emerging as a potential solution to overcome the drawback of the conventional inertial measurement sensors with a soft, flexible design that allows for direct body-conforming adhesion to the skin. Thus, this study aims to establish the viability of the flexible inertial sensors for the ambulatory measurement of human knee flexion-extension (F/E) during level walking. The accuracy of flexible inertial sensors with our proposed algorithm is investigated by comparing the derived kinematics with those measured using a conventional motion capture system. Three volunteers (two males and one female, aged 24-27) served as subjects. Three flexible inertial sensors were placed on the subjects' right legs, two on the thigh (lateral thigh and distal anterolateral thigh), and one on the lateral shank, and subjects performed level walking with varied walking cadence (80, 100, and 120 steps/min). The overall root-mean-square-error across all participants is 7.8° for distal anterolateral thigh sensor (using the data of distal anterolateral thigh sensor and shank sensor to estimate knee F/E) and 4.5° for lateral thigh sensor (using the data of lateral thigh sensor and shank sensor to estimate knee F/E). The results illustrate that our proposed method can generate comparable results with motion capture systems and is insensitive to different walking cadence. A significant effect of sensor placement is observed, and a more proximal lateral thigh strategy is suggested. The advantages of the flexible inertial sensors in human motion analysis over the existing inertial measurement units are also highlighted.