With more attention on personal privacy and the need for a security defense, it is necessary to design an intelligent lock system with a higher security performance. Here, a novel high security double lock system integrating triboelectric nanogenerators (TENGs) with a double bubble structure (DB-TENG) and deep learning models is proposed. The TENG as a self-powered sensor is developed using silicone rubber and copper foil. By optimizing the thickness of the top layer film, surface microstructure, the size of the air bubble, and design of the double bubble structure, the sensitivity of the DB-TENG reaches 19.08 V/kPa. For the feasibility study, the sensor is fabricated to a smart belt to collect respiratory behaviors as a respiratory code. A Long Short-Term Memory network is adopted to identify four typical respiratory signals with an average accuracy of 97.00%. The system is deployed on a Raspberry Pi to determine whether the user is permitted through both the collected respiratory code and the related face image and will send an alarm message if one of the two does not match. It is worth mentioning that users can send alarm signals undiscovered by controlling their respiratory signals. Therefore, the proposed system has superb potential in security demanding environments.