There is provided a method for predicting the perceptual quality of a signal (310). The method comprises inter alia subjecting the signal (310) to a plurality of objective quality measurement algorithms, and forming a plurality of normalized weighted sums (330a-e) based on the output (320a-c) of the objective quality measurement algorithms. The weights used to form the normalized weighted sums (330a-e) are such that each of the normalized weighted sums (330a-e) gives an optimal prediction of the perceptual quality of signals for an associated target value. The normalized weighted sums (330a-e) are then weighted based on their distances to their associated target values. The method provides an improved prediction accuracy of the perceptual quality.