Abstract: In the context of “New Engineering”, targeting the experimental teaching for automation majors, a comprehensive and innovative experiment on a speech-recognition-based forearm manipulator control system is designed. This experiment achieves control of the forearm manipulator through speech recognition, utilizing the STM32F405 as the voice control chip. The system employs the MPU6050 to measure the three-axis angular velocity and three-axis acceleration of the arm. The sensor data is processed and analyzed using a gradient descent algorithm based on quaternions to determine the real-time pose of the forearm manipulator. The motors are controlled to maintain its hand level with the horizontal plane during movement, thereby achieving self-balancing. The system utilizes an OpenMV machine vision module to perform grip correction when the manipulator executes grasping tasks, thereby achieving closed-loop control. It has been demonstrated that implementing this speech-recognition-based embedded control system experiment for manipulators in universities meets the needs of modern experimental teaching, helping students to better understand and apply the construction process of embedded platforms and to skillfully handle various complex engineering problems.