Implementing Speech Feature for Embedded System to Support Color Blind People

Abstract

Speech feature can be an alternative interaction model for embedded device that is designed without many buttons for control. In this work, we create colorblind aid system. This paper compares speech feature created using SAPI5.1 with speech feature created manually using phoneme concatenation and DTW, and assess their performance in WinCE6 and WES09 embedded OS environment. The testing for speech feature is conducted by 10 respondents to identify words synthesized by systems and to say words to be recognized by systems. Testing for the OS is conducted by observing size, boot time, and their support for the speech feature. As the result, speech feature created using SAPI5.1 is better than the manually-created one, with success rate 88, 33% for speech synthesis and 75, 87% and 74, 76% subsequently for speech recognition in silent and noisy condition. The second test shows that WES09 is more suitable for this particular system due to its support for .NET and SAPI5.1.