Content-based authentication using digital speech data

Abstract EN

A watermarking technique for speech content and speaker authentication scheme, which is based on using abstracts of speech features relevant to semantic meaning and combined with an ID for the speaker, is proposed in this paper.

The ID which, represents the watermark for the speaker, is embedded using spread spectrum technique.

While the extracted abstracts of speech features are used to represents the watermark for the speech, is embedded in the original speech file using secret key.

The abstracts speech feature is implemented using B-spline curve interpolation.

The paper provides background knowledge for the concept of speaker watermarking and content-fragile watermarking based on digital speech data.

Then, the suggested feature based authentication scheme is develop and the results from the evaluation are presented.

It shows that the suggested scheme is successful for combining speech and speaker watermark authentication.