The pooled scalogram: A wavelet-based approach to detect the co-activation of several muscles in the time-frequency domain

Objective: Different approaches based on surface electromyographic signals (EMG) were proposed to understand the interplay among several muscles, most of which were based on the solely EMG amplitude and neglected spectral characteristics. This study introduced an approach that quantitatively characterizes the co-activation of multiple muscles in time-frequency domain, based on the continuous wavelet transform (CWT). Methods: The pooled scalogram, defined here as the CWT-based cross-energy localization of multiple EMG signals, quantified the co-activation among several muscles in time-frequency domain. Algorithm performances were tested on 30,000 synthetic EMG signals, with different signal-to-noise ratio. Experimental sEMG were recorded from tibialis anterior, gastrocnemius lateralis, and vastus lateralis during walking of 31 young healthy subjects. Significant invariants were extracted by Non-Negative Matrix Factorization Analysis to determine timefrequency invariants. Results: The proposed CWT approach provided an accurate prediction of co-activation timing in the synthetic signal. Analysis of experimental signals indicated population consistency in time-frequency invariants and showed that muscular co-contraction peak was identified in heel strike and in the Piper frequency band. Conclusion: Synthetic data demonstrated the capability of the pooled scalogram to accurately detect the timings among multiple signals. Experimental analysis suggested that when TA, GL, and VL are co-active over time, they also synchronized in frequency within the Piper band. Significance: The pooled scalogram will lead to a better understanding of the control laws underlying the motor coordination, and it could represent a marker of neurological pathologies or a quantitative evaluation of the effects of a rehabilitative process.