Cisco Technology, Inc.

Avaya Inc.

Wai-Yip Chan

Jiandong Shen

Tiago H Falk

Qingfeng Xu

A non-intrusive speech quality estimation technique is based on statistical or probability models such as Gaussian Mixture Models ('GMMs'). Perceptual features are extracted from the received speech signal and assessed by an artificial reference model formed using statistical models. The models characterize the statistical behavior of speech features. Consistency measures between the input speech features and the models are calculated to form indicators of speech quality. The consistency values are mapped to a speech quality score using a mapping optimized using machine learning algorithms, such as Multivariate Adaptive Regression Splines ('MARS'). The technique provides competitive or better quality estimates relative to known techniques while having lower computational complexity.

Single-sided speech quality measurement

The present invention relates to systems and methods for compressing, decompressing, and transmitting video data. The systems and methods include pixel by pixel motion estimation and compensation and efficient quantization of residual errors. The present invention applies block estimation of the residual error produced by motion compensation. The block estimation is applied by a local decoder to generate synthesized blocks of video data. The block estimation approximated uses a set of predetermined motion estimation errors that are stored as error vectors in a codebook. The codebook is included in an encoder of the present invention and converts an error vector for each block to an error vector index. The error vector index, which introduces minimal transmission burden, is then sent from the encoder to a target decoder. A receiving decoder also includes a copy of the codebook and converts the error vector index to its associated error vector for reconstruction of video data.

Growing community of inventors

Kingston, Canada

Wai-Yip Chan