The assessment of local muscle fatigue during dynamic exercises is viewed as essential, since people tend to maximise effective muscle training and minimise harmful muscle injuries. This study aims to investigate how surface electromyography (sEMG) and subjective metrics together can play a role in dynamic muscle fatigue prediction during exercises to promote the design of health and fitness technology. 20 healthy male participants were recruited in the experiment, and sEMG and self-reported data were collected. Features in temporal and spatial domain were extracted from sEMG data, such as RMS (root mean square) and FInsm5 (spectral parameter proposed by Dimitrov). Our results showed that some sEMG features indicated directional changes with the increase of dynamic muscle fatigue. Spearman correlation analysis indicated that Borg ratings had strong correlations with RMS and FInsm5 (spectral parameter proposed by Dimitrov) slopes. Then this paper discusses how to use sEMG and Borg data to evaluate muscle fatigue during exercises. A framework is proposed based on the joint analysis of spectra and amplitudes across RMS and FInsm5 slopes. This paper further discusses how to design health and fitness technology for the benefits and the limitations of the study.