Partial discharge detection is critical for ensuring the stable operation of Gas Insulated Switchgear (GIS) equipment. This paper addresses the issue of partial discharge detection in GIS by conducting research on the acquisition and analysis of discharge acoustic signals using MEMS microphones. Firstly, four typical discharge models commonly found in GIS were constructed, and the corresponding discharge signals were collected using MEMS sensors. Through time-domain and frequency-domain analysis of the acquired signals, multiple feature values were extracted, including traditional statistical features and newly introduced periodic features. Subsequently, a Support Vector Machine (SVM) classifier was employed to classify and identify the extracted features. Experimental results demonstrate that the classification accuracy significantly improved with the inclusion of periodic features compared to traditional methods.