The paper is quite interesting and timely.
The methodology proposed are not fully new, the use of CWT is appropriate and the Morse wavelet is originally applied, as far as I know.
I appreciate the idea behind Figure 5, however it is not completely readable, I suggest to improve tha caption and make an additional figure that focuses on the main aspects of interest that can be read from the figure.
Some errors and typos are present all along the text. Please use First instead of Firstly; p.6, Mortlet is Morlet; there are no Acknowledgments, if not cancel the Section title. The concept of validation patience (p.8) is not fully clarified, please improve the description (is a Figure useful?).
You make validation of your model with k-fold (k=10) cross-validation; it could be interesting to make a testing on out-of sample fresh data, also 1 or 2.
With regard to the achieved results, add a note saying that the comparison is made on different datasets.
The comparison of WT with STFT is somewhat well known to the prospective readers, please better focus on the fact that STFT gives a coverage of the time-frequency space which is uniform, in contrast to CWT, and argue on the impact of this aspect. Please also make reference to the use of a specific windowing on CWT, as a difference with STFT.
With regard to the difference of Morlet waelet with your approach and with Mexican Hat, I suggest to make reference to the following paper, although it refers mainly to EMG signals:
A. Greco, D. Costantino, F.C. Morabito, M. Versaci, A Morlet wavelet classification technique for ICA filtered sEMG experimental data, Proceedings of the IJCNN 2003, 2003, pp. 166-171, ISBN: 1098-7576.