This study proposes a non-invasive smart meter system that considers the power use habits of users unfamiliar with electric appliances, and can be used by inserting the smart meter into an electrical circuit. This study also creates a database mechanism, appliance recognition classification, and a waveform recognition method, in order to solve the large data volume problem in current appliance recognition systems. In comparison to other appliance recognition systems, the low-end embedded system chip used in this study has low power consumption, as well as high expandability and ease of use. This experiment is different from the research environments of other appliance recognition systems by considering parallel multi-appliance recognition and general users' habit of using power. This study will not make any assumption of power utilization in the experiment. The total system recognition rate is 84.42%, and the total recognition rate of a single electric appliance is 93.82%, proving the high feasibility of this study.