Implementation Of Speaker Recognition Using Support Vector Machines For Security Systems


Creative Commons License

Özkaynak E. , Sulak E. E. , Yılmaz H.

8. International Advanced Technologies Symposium, Elazığ, Türkiye, 19 - 22 Ekim 2017

  • Basıldığı Şehir: Elazığ
  • Basıldığı Ülke: Türkiye

Özet

Nowadays, hand in hand with the developments in technology, various methods have begun to be developed in the field of communication between people and machines. Given the work done, it seems that voice communication between machines and people has recently become more important. Especially in this period where security systems are important in every area, voice recognition systems have become one of the subjects that are working intensively in this field. Voice recognition studies are categorized as speech recognition, speaker recognition, identification, and gender determination within themselves. When examining the processing steps of the voice recognition process: first, the voice is taken in the appropriate environment and with appropriate equipment, then the voice is detected, after removing the noise, the necessary filtering is done to detect the beginning and end of the voice, the voice is compared and the results are evaluated. Obtaining the most accurate and reliable result by using appropriate methods of all operations to be performed for voice recognition is the most important element of voice recognition systems. After performing the necessary preliminary actions on the voice data, the results of voice recognition, in other words, recognition of a security system, are evaluated using the Support Vector Machines method. 10 voice data from each person who are 4 males and 4 females in total of 80 voice data were used for this study. As a result of this, Support Vector Machines have achieved a rate of 86.66% voice recognition for males and 80.18% for females.