Low-Power Microcontroller Implementation of a Voice Command Interface for IoT Nodes

This paper describes a low-power voice-activated prototype providing a hands-free user interface for voice commands. Utilizing a microelectromechanical-system microphone, the device extracts frequency-domain voice features to discern different keywords, enabling its use in IoT applications for medical support or office assistance. The keyword spotting functionality is implemented on an ARM Cortex-M4 microcontroller. Data acquisition is facilitated by a 12 -bit built-in analog-to-digital converter. Mel-frequency cepstral coefficients are calculated to create a compact representation of voice data. Dynamic time warping is used for feature comparison. The keyword spotting algorithm demonstrates 95% accuracy. By reducing power consumption, this research leads to the advancement of IoT technology for applications that need cloudless computing and hands-free interfaces.