During a sustained contraction the surface my oelectric signal undergoes amplitude scaling (by a factor h), due to reduction or loss of motor unit B, dinn potentials, and time scaling (hy a factor k) due to reduction of muscle fiber CV and other fac tors. These changes are particularly evident in the responses evoked by electrical stimulation (M-waves) when stimulation rate is above 30 or 35 Hz. The time expansion factor corresponds to a frequency com pression factor of the signal's power spectral density (PSD) function and in previous research it has been estimated from the progressive decrease of the mean and median frequencies (MNF and MDF) of the PSD function.Consider the short-term-stationary signals :Cl(t) and :l: 2 (t) = h:l:l(kt). H :l:l(t) has PSD Pl(f), then :l: 2 (t) has PSD given by P2(f) = h 2 /k 2 • P(f Ik). The MNF and MDF of the PSD Pl(f) are defined as:It can be shown that hmed = k hmed and hmea.n = khmean. and that A 2 = A1hlk (where Al and A2 are the average rectified values of :l: l (t) and :C2(t) over the time interval 0 -T) from which an estimate of Ie and h may be obtained. This method does not provide an indication of how" good" the estimates of k and h may be. If ;J)2(t) is not an approximatly scaled version of Xl(t), the estimates of k and h, based on spectral variables, are poor. In particular, this is often the case during a sustained contraction when stimula tion is applied at rates above 30 Hz and the response evoked by one stimulus may be truncated by the next stimulus generating spurious high frequency compo nents in the PSD resulting in the overestimation of MDF and MNF. 0-7803-0785-2/92$03.00 ©IEEE An alternative way to estimate k and h is to scale :l:l(t) in amplitude and time until a "best" match with :l: 2 (t) is obtained. IT a)2(t) is truncated, only the portion of:l:1(1 ) common to :l: 2 (t) should be used. For sampled signals, time sca.ling implies misalignement of the samples of :l:l(t) and :l: 2 (t ) . Interpolation is therefore necessary to compute the error between the two signals for the same t values. In this work we use cubic splines for the interpolation of the scaled signal and the LMS criterion to define the error e 2 which we minimize by finding the k and h pair that nulls its gradient 'h,lt:. ---Oe 2 (n) Be2(n) V'h,lt:e2(n) = 0 lor Oh = 0 and Ok = 0where Xl (kn) is a scaled and interpolated version of :1:1 (n) re-sampled at na nd:

In order to better assist the rehabilitation treatment of patients with musculoskeletal injury, standard rehabilitation actions are needed to guide the musculoskeletal rehabilitation process. With more and more urgent demands, the musculoskeletal rehabilitation evaluation systems have attracted a high degree of attention. Experts have proposed a series of systems based on laser, ultrasound and image, which can give reasonable recognition and judgment. However, these systems either require specialized and expensive equipment or can be affected by ionizing radiation. How to construct a musculoskeletal rehabilitation evaluation system with low cost, good effect and little injury is still a great challenge. In this paper, we propose MSEva, a musculoskeletal rehabilitation evaluation system based on EMG signals. Specifically, the system uses EMG sensor to collect a large amount of data for 5 rehabilitation actions. Secondly MSEva uses Wavelet Transform (WT) to extract the signal features, and then puts the processed data into the Long Short-Term Memory (LSTM) network for model training. Finally, the system uses LSTM model to evaluate the normality of the EMG response of rehabilitation actions. The results show that the average accuracy of MSEva reaches 94.37%, which has important evaluation value in guiding the rehabilitation of musculoskeletal patients.

No abstract