#TITLE_ALTERNATIVE#

• Main Author: NUGRAHA (NIM 13203075), MAHESA
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/10835
• Institution: Institut Teknologi Bandung

Speech recognition has been developed for numerous applications with primary focus have been done on desktop computer and digital signal processor (DSP). However, some applications require implementation on a low-cost and lowpower system. These applications prefer a microcontroller based implementation on account of cost, power, and portability. Microcontroller based implementation requires a need of simpler algorithm than desktop computer based implementation. This need is based on microcontroller constraint such as computational power and memory.<p> <br /> <br /> <br /> <br /> <br /> <br /> A speech recognition system using formant tracker has been designed for this final project. On this algorithm, the first two formant frequency of a speech signal is extracted into a feature by detecting the sum of zero-crossing in the processed speech signal. This algorithm does not require intensive computational process and can be efficiently implemented on 8-bit microcontroller.<p> <br /> <br /> <br /> <br /> <br /> <br /> System implementation is done on personal computer (PC) with MATLAB and on microcontroller. Maximum average recognition rate of the system reached 87,5% on MATLAB and 85% on microcontroller.