Voice classification are often classified by using Recurrent Neural Network and 1-Dimensional Convolutional Neural Network (1D-CNN). 1D-CNN is used for classification model. In general, 1D-CNN learns original data. However, in this study, original data is replaced with surrogate data. Test accuracies at that time are compared. In this way, we search which part of time series data is effective for using 1D-CNN