Vector quantization (VQ) is a well-known lossy compression method, which has not often been applied to biosignals. In this paper, VQ and its mean residual variant for encoding and decoding electromyography (EMG) signals have been tested. The methods are selected in such a way that they can be later applied in a low-resource embedded system. A neural network approach is used for codebook generation. The preservation of medical parameters is a prominent sign of quality in medical compression systems. Both signal level fidelity factors and preserving medical parameters are tested. The results show that mean residual vector quantization with short segments is a workable approach for EMG signal compression.