Key points are not available for this paper at this time.
In the ever-evolving landscape of technology, the integration of intelligent and responsive devices into our daily lives has become ubiquitous. From smart homes to industrial settings, the demand for efficient and realtime processing at the edge of networks has spurred the development of compact and powerful solutions. Studies show that one such paradigm-shifting advancement is the emergence of Tiny Machine Learning (TinyML), which empowers edge devices with the capability to execute machine learning (ML) algorithms on constrained hardware. The study addresses the challenge of balancing computational capabilities with energy consumption in voice-enabled edge devices. This paper also explores the integration of TinyML, particularly Convolutional Neural Network (CNN), with low-powered microcontrollers for voice-controlled edge devices. Using on a smart fan as a case study, the methodology involves data collection, CNN model design and training, model optimization, and deployment on an Arduino Nano 33 BLE Sense microcontroller. Experimental results indicate a 95% accuracy in recognizing voice commands while controlling the smart fan, showcasing the potential of TinyML for efficient and responsive edge computing applications.
Uddin et al. (Wed,) studied this question.