1: \begin{abstract}

2: Electromyogram (EMG) classification is a key technique in EMG-based control systems. 

3: The existing EMG classification methods do not consider the characteristics of EMG features 

4: that the distribution has skewness and kurtosis, 

5: causing drawbacks such as the requirement of hyperparameter tuning. 

6: In this paper, we propose a neural network based on the Johnson $S_\mathrm{U}$ translation system 

7: that is capable of representing distributions with skewness and kurtosis.

8: The Johnson system is a normalizing translation 

9: that transforms non-normal data to a normal distribution, 

10: thereby enabling the representation of a wide range of distributions.

11: In this study, a discriminative model based on the multivariate Johnson $S_\mathrm{U}$ translation system 

12: is transformed into a linear combination of coefficients and input vectors using log-linearization. 

13: This is then incorporated into a neural network structure, 

14: thereby allowing the calculation of the posterior probability of the input vectors for each class and

15: the determination of model parameters as weight coefficients of the network. 

16: The uniqueness of convergence of the network learning is theoretically guaranteed.

17: In the experiments, the suitability of the proposed network for distributions including skewness and kurtosis 

18: is evaluated using artificially generated data. 

19: Its applicability for real biological data is also evaluated via an EMG classification experiment. 

20: The results show that the proposed network achieves high classification performance without the need for hyperparameter optimization. 

21: \end{abstract}

22: