Muscle fatigue in human lifting motion has been found to be one key factor of repetitive stress injury (RSI). Investigating muscle fatigue in this motion could provide valuable clues in designing and controlling the wearable device to reduce the likelihood of human RSI. Therefore, the main goal of this study is to quantify the effects of muscle fatigue on synergy during human lifting motion. Eight healthy adults complete the lifting tasks until extremely exhausted. Firstly, the surface Electromyography (sEMG) signals are analyzed using root mean square and continuous wavelet transform to demonstrate fatigue. The variance accounted for (VAF) approach is then used to extract three muscle synergies. Finally, the pre-processed signals are decomposed by a non-negative matrix decomposition algorithm into muscle synergy relative weights and muscle activation coefficients. The obtained results reveal that despite almost remaining unchanged before and after fatigue in terms of synergy structures, the trend of relative weights and the time for maximum activation of muscle synergy to occur both have been discovered to be different. This discovery could imply that human beings alter our muscle synergy strategies to compensate for impacts of local muscle fatigue in different ways under lifting motion, which would provide meritorious insights on designing and controlling the assistive robotic device.