A Low Cost General Purpose Speech Recognition System Used to Control Processes in The Embedded and Stationary System and Write Texts on The PC

Creative Commons License

Burunkaya M., Dijle M.

JOURNAL OF POLYTECHNIC-POLITEKNIK DERGISI, vol.21, pp.477-488, 2018 (ESCI) identifier


In this study, a low cost, microcontroller based speaker dependent speech recognition system was designed and parameters affecting speech recognition with high accuracy were investigated and a speech control system which can be used as a general purpose was obtained. In the realized system, the relevant data is transferred to a computer (PC) in order to write text. For these purposes, the system can be used by people with disabilities or healthy people. Using the speech recognition circuit, the letters of the alphabet and some selected words for control operations are recorded in advance in this system which can be trained. Speech voices detected by the microphone in the system are converted into digital signals corresponding to them. Once the data for the recognized voices are compared using a microcontroller, they are transferred to the PC using RS232 and PS2 communication protocol. The flexible structure of the system and the ability to be used stand-alone without PC when necessary also improves its usability. The system has been tested in noiseless environment and with different noise levels, with letters, homophones and non-homophones words. The obtained data were analyzed statistically by using SPSS package program. During the experiments, Rion NL-21 instrument was used for noise measurements.