Signal restoration using left-sided and right-sided autoregressive parameters
AssigneeLucent Technologies Inc.
InventorWalter Etter
Life of this patent
Estimating the term…
Abstract
Signal reconstruction methods reconstruct a missing portion of a signal from a first known portion of the signal preceding the missing portion, and a second known portion of the signal succeeding the missing portion. The first known portion of the signal is represented using a first autoregressive model and the second known portion of the signal is represented using a second autoregressive model. These methods construct a signal estimate from the first and second autoregressive models, thereby permitting the reconstruction of any signal having a missing portion, irrespective of whether or not the length of the missing portion is greater than the short-term stationarity duration of the signal.
Claims23
A signal reconstruction method for reconstructing a missing portion of a signal, the method comprising the steps of:
- (a) generating a first estimate of the missing portion of the signal from a first known portion of the signal preceding the missing portion and a second known portion of the signal succeeding the missing portion;
- and
- (b) generating a new estimate of the missing portion from the first estimate, from the first known portion of the signal, and from the second known portion of the signal.
A signal reconstruction method for reconstructing a missing portion of a signal from a first known portion of the signal preceding the missing portion, and a second known portion of the signal succeeding the missing portion, wherein the first known portion of the signal includes a first set of known samples, the second known portion of the signal includes a second set of known samples, and the missing portion of the signal is reconstructed by predicting values for a set of missing samples using the steps of:
- (a) representing the first set of known samples using a first autoregressive model and representing the second set of known samples using a second autoregressive model;
- and
- (b) generating initial estimates for the missing samples based upon the first autoregressive model and the second autoregressive model.
A signal reconstruction system for restoring a missing portion of a digitized signal based upon a first known portion of the digitized signal and a second known portion of the digitized signal, the digitized signal including a plurality of samples, the system including:
- (a) a missing sample detector for determining whether or not any samples are missing from the digitized signal;
- b) a delay line for applying a predetermined amount of time delay to the digitized signal;
- (c) an interpolator for estimating the missing portion of the signal from the first known portion and the second known portion of the digital signal;
- (d) a switch, coupled to the missing sample detector, for directing the digitized signal to the delay line if no samples are missing from the digitized source signal;
- otherwise, directing the digitized signal to the interpolator;
- the interpolator comprising:
- (i) a segmentation device for determining the first known portion of the digital signal and the second known portion of the digital signal, and
- (ii) a segmentation parameter estimator, coupled to the segmentation device, for estimating parameters N.sub.L, N.sub.R, and M, wherein N.sub.L, represents the number of samples in the first known portion, N.sub.R represents the number of samples in the second known portion, and M represents the number of samples in the missing segment;
- the segmentation parameter generating a left-sided sample vector in response to the first known portion and a right-sided sample vector in response to the second known portion;
- (iii) a vector augmentation device, coupled to the segmentation parameter estimator, for generating K augmented left-sided sample vectors and K augmented right-sided sample vectors, wherein K is a positive integer;
- (iv) an autoregressive parameter estimator coupled to the vector augmentation device for determining left-sided autoregressive parameters for the augmented left-sided sample vectors and right-sided autoregressive parameters for the augmented right-sided sample vectors, the left-sided autoregressive parameters denoted by a vector .alpha..sup.(k), and the right-sided autoregressive parameters denoted by a vector b.sup.(k),
- (v) a matrix inverter, coupled to the autoregressive parameter estimator and to the segmentation device, for determining an estimated signal for the missing segment, denoted as vector x.sup.(k), in response to vector .alpha..sup.(k) vector b.sup.(k), and the segmentation device.
Dependencies are inferred from the claim text.