Voice Command Smart Home Control on Single-board Computer Using Google Speech API and Cosine Similarity

Abstract

The development of technology today is very fast and sophisticated. One example of technological developments is the smart home. A smart home is a house or building equipped with high technology that allows various systems and devices in the home to communicate with one another. This technology is very helpful for parents and people with disabilities in activities at home. One way of controlling smart homes currently being developed is voice control. However, some smart home devices do not yet have control features using voice commands in Indonesian. This study aims to create and test a smart home device that can be controlled by voice commands using the Google Speech-to-Text Application Programming Interface (API) as a medium for capturing and converting voice commands from users into text. The results of the conversion are processed using the Cosine Similarity method to determine what commands are given and which devices the user wants to control. This process is carried out on a Single Board Computer (SBC), specifically the Raspberry Pi. The system achieves up to 100% accuracy in quiet conditions with an average response time of 1001.4 ms. However, this accuracy drops to approximately 80% in noisy environments above 65dB, indicating limitations of the Google Speech API under high noise conditions.