In recent years, electromyogram signal (EMG) feature selection, based on wavelet transform, has received considerable attention. This study introduces a new multi-wavelet function for surface EMG (sEMG) signal intended for tasks that involve hand movement recognition. To create the new wavelet function, several types of well known mother wavelet were exploited and through their integration the proposed mother wavelet was generated. The proposed wavelet function closely reproduced the characteristics of the EMG signal, while increasing the recognition accuracy of hand movements. We used eight unique classes of hand motions and considered the ability of various mother wavelets and the proposed multi-wavelet to recognize these movements. Furthermore, we used local extrema and zero crossing (ZC) as DWT features. The results demonstrate that the proposed multi-wavelet function provides 87% recognition mark compared to 78%, the best performance that any other mother wavelet was able to achieve.