The system assists in detecting people trapped in cars using sound recognition technology to improve traffic safety efficiency, especially in emergency situations. The system aims to detect the presence of a person trapped in a vehicle through the analysis of characteristic audible signals such as cries for help, children's voices, knocking on doors, or sounds caused by a car accident. The system works by picking up sound in the cabin through pre-installed microphones. These audio signals are then analyzed using machine learning algorithms to identify unusual sounds, such as screaming, banging on doors, or knocking hard objects on vehicle parts. When these sounds are detected, the system will automatically activate a warning to the driver or notify the rescue agencies for timely handling measures. The research content uses the sound recognition method using Machine Learning/AI Model and Matlab software to simulate the operation of the system.