High-rate data embedding in unvoiced speech.

摘要: We propose a blind speech watermarking algorithm which allows high-rate embedding of digital side information into signals. exploit the fact that well-known LPC vocoder works very well for unvoiced speech. Using an auto-correlation based pitch tracking algorithm, voiced/unvoiced segmentation is carried out. In segments, linear prediction residual replaced by data sequence. This substitution does not cause perceptual degradation as long residual’s power matched. The signal resynthesised using unmodified filter coefficients. watermark decoded analysis received and extracted from sign residual. nearly imperceptible provides channel capacity up to 2000 bit/s in 8 kHzsampled signal.