Abstract <p>Assessment of lower limb spasticity is crucial for enhancing both functional recovery and rehabilitation effectiveness of stroke patients. Traditional methods like the Modified Ashworth Scale (MAS) are often hampered by subjective bias, leading to inconsistent assessments among clinicians. In contrast, neurophysiological methods, mainly surface Electromyography (sEMG) data analysis, offer a more objective and sensitive alternative. Most of the existing studies based on sEMG focus more on upper limb applications, while research on the assessment of lower limb spasms is still under exploration. In this study, a feasibility investigation was conducted to explore the quantitative assessment of post-stroke lower limb spasticity based on sEMG features. The research aimed to elucidate the potential association between spasticity severity and sEMG features through sEMG analysis. Eleven participants with stroke were recruited and stratified into three MAS levels. Lower limb muscle activity was recorded using sEMG sensors during spasticity assessments performed at three stretching velocities (slow, medium, and fast). sEMG features were extracted to systematically analyze correlations with spasticity grades, following which appropriate features were selected. Machine learning algorithms were subsequently employed to validate the feasibility of sEMG-based quantification for lower limb spasticity classification. Experimental results demonstrated the classification accuracy of 86.17% and the F1-score of 86.17% in spasticity evaluation utilizing sEMG features. This study provides a new research idea for the relationship between sEMG-derived features and spasticity grading, while enhancing the mechanistic understanding of sEMG features contributions to spasticity pathophysiology.</p> Graphical abstract <p></p>