Segmentation of speakers during emergency call with watermarking technique

The article presents test algorithm that is used for segmentation of the speakers in a telephone conversation. The proposed algorithm is based on the use of a watermark inserted into the speech signal. The program was written in the Matlab / Simulink environment and implemented on a signal processor TMS320C6713 DSK. A watermark insertion algorithm was tested, and the signal from the telephone line was recorded using the prepared phone-to-PC adapter. In addition, an algorithm used to detect the watermark from the previously recorded conversation was created, in order to separate the speakers: an operator receiving the call and a citizen calling to the emergency services.