| 摘要 |
The present invention relates to a method and apparatus for deconvolving a noisy measured signal obtained from a sensor device (100), said noisy measured signal (y(t)) being the sum of the convolution product ( x(t) ? N(t) ) of an input signal (x(t)) of the sensor device, representative of a feature of physical quantity, by a convolution kernel (N(t)) defined by the response function of the sensor device (100) and a noise which interfere with the measure. The method is characterised in that said method comprises an estimate computation step (400) in the course of which a minimal estimate (x min (t) ? N(t) ) of the convolution product of the input signal by the convolution kernel of the sensor device is computed in order that said minimal estimate stays below the noisy measured signal (y(t)) and has at least one point in common with the noisy measured signal (y(t)). |
