Abstract: In the context of “New engineering”, for experimental teaching in automation majors, integrating cutting-edge technology of science and technology, a comprehensive and innovative experiment of “Speech Recognition Based Forearm Manipulator Control System” is designed. Using STM32F405 as the core chip, the device can control the manipulator through the speech recognition module. The system uses the MPU6050 to measure the three-axis angular velocity and three-axis acceleration of the arm, and uses the Gradient descent based on Quaternion to process the sensor data to obtain the immediate attitude angle of the forearm manipulator. In the process of movement, the manipulator motor is controlled to maintain its hand level with the horizontal plane when moving, achieving self-balance. Using the OpenMV module, the system performs grasping calibration during the robotic arm’s grasping task, achieving closed-loop control. Experiments on embedded control systems for speech-recognition-based manipulators in universities meet modern teaching needs, helping students understand and apply embedded platform development while solving complex engineering problems.