In this study, we aim to develop a bioacoustics-based monitoring system for non-native species such as the Tokyo daruma pond frog (Pelophylax porous porous) in Hokkaido. We created a sound source separation model based on “Cocktail Party Source Separation” to separate frog calls from mixed audio recordings. A neural network model was trained using data from the Japanese tree frogs (Hyla japonica) and Black-spotted pond frogs (Pelophylax nigromaculatus) to separate mixed frog calls. The evaluation results showed successful separation with some issues in frequency-range after source separation. Future improvements are required to refine the model by incorporating multi-channel data for better separation quality to improve the performance of species identification and sound source estimation.