| 摘要 |
The method of estimation of a relationship between the control parameters of a system for the compression of digital data representative of physical quantities and the quantities characterizing the system performance includes an estimation of the parameters of parametric affine model by adaptive filtering. The breakdown of a model is detected by measurement of filter innovations and the detection of variations of the innovations. The model parameters are compensated when the variations are greater than a predetermined threshold. The adaptive filtering is a Recursive Least Square (RLS) filtering. The detection and the compensation are effected by use of the Generalized Likelihood Ratio (GLR). The system quantities include the rate of coded data, and the control parameters are determined for the system providing a predetermined rate or distortion. The system quantities include the distortion of coded data, and the control parameters are determined for the system providing predetermined rate of distortion. A device for an estimation of the relationship between the control parameters and the quantities characterizing the system performance comprises the means for the estimation of parameters of parametric affine model by adaptive filtering, the means for the detection of breakdown of model by measuring the filter innovations and the detection of variations of the innovations, and the means for the compensation of model parameters when the variations are greater than a predetermined threshold. The means for estimation, detection and compensation include a microprocessor, e.g. central processing unit (CPU), a read-only memory (ROM) storing a program for signal compression, and a random access memory (RAM) containing data registers adapted to register the variables modified in the course of program execution. Independent claims are included for an estimation device, and for a digital device comprising the estimation device.
|
