ABSTRACT
This paper presents an Automatic Speech Recognition (ASR) system that can be efficiently used for personalized lighting control. Using a Deep Neural Network-Hidden Markov Model (DNN-HMM) algorithm, the system implements the dimming, color temperature and color adjustment of Light-Emitting Diodes (LEDs) according to 31 sets of human command words. Tests were conducted at distances of 20 cm, 50 cm and 100 cm in quiet and noisy environments. Results show that the mean accuracy of speech recognition is 94% and the response time is 0.7 s, which indicates good potential for commercialization.
