This paper proposes the feasibility of a stiffness measurement for muscle contraction force estimation under muscle fatigue conditions. Bioelectric signals have been widely studied for the estimation of the contraction force for physical human-robot interactions, but the correlation between the biosignal and actual motion is decreased under fatigue conditions. Muscle stiffness could be a useful contraction force estimator under fatigue conditions because it measures the same physical quantity as the muscle contraction that generates the force. Electromyography (EMG), mechanomyography (MMG), and a piezoelectric resonance-based active muscle stiffness sensor were used to analyze the biceps brachii under isometric muscle fatigue conditions with reference force sensors at the end of the joint. Compared to EMG and MMG, the change in the stiffness signal was smaller (p < 0.05) in the invariable contraction force generation test until failure. In addition, in the various contraction level force generation tests, the stiffness signal under the fatigue condition changed <10% (p < 0.05) compared with the signal under non-fatigue conditions. This result indicates that the muscle stiffness signal is less sensitive to muscle fatigue than other biosignals. This investigation provides insights into methods of monitoring and compensating for muscle fatigue.