Dataset for: Multimodal Upper Limb Rehabilitation Assessment System integrating Mechanomyographic and Surface Electromyographic Signals

Purpose: Objective and accurate clinical muscle strength assessment is the foun-
dation for formulating personalized rehabilitation programs for stroke survivors.
However, traditional scales are mostly limited by the subjectivity of clinicians.
This study proposes a quantitative assessment framework integrating surface
electromyography (sEMG) and mechanomyography (MMG).
Methods: This study collected multimodal signals from the deltoid muscle.
High-dimensional features were extracted from the time domain, frequency
domain, and cepstral domain, with the innovative introduction of Mel Frequency
Cepstral Coefficients (MFCC) to capture fine mechanical vibrations in MMG sig-
nals. Five commonly used machine learning models were constructed based on
multimodal data and individual MMG data respectively, including Support Vec-
tor Machine (SVM), Random Forest (RF), Decision Tree, Logistic Regression,
and Gradient Boosting Tree. The model performance was evaluated using four
metrics: Accuracy, Precision, Recall, and F1 Score. Feature importance analy-
sis was conducted for both data approaches to elucidate the electro-mechanical
signal fusion mechanism。Results: Among the five evaluated machine learning models, the RF model
achieved the best performance with an F1 score of 0.9279, representing a 27.14%
improvement compared to the single-modal electromyography model. The F1
scores of the Logistic Regression, Decision Tree, SVM, and Gradient Boost-
ing Tree models were 0.8139, 0.8820, 0.9035, and 0.9148 respectively. Feature
importance analysis revealed that, through feature weight transformation, the
relationship between sEMG and MMG is not a simple linear mapping but exhibits
complex nonlinear dynamic characteristics.
Conclusion: The integration of sEMG and MMG signals effectively enhances
the accuracy of rehabilitation assessments. This non-invasive, high-precision
technology offers a novel rehabilitation assessment method for clinical settings,
presenting new feasibility for objective rehabilitation evaluation in both clinical
and remote rehabilitation environments