Because of the increasing demand of low power digital systems, it is of great interest to extend the existing high-level power estimation techniques to handle flexible data models, as they appear in relevant applications. This paper presents a data model and an algorithm suitable for estimating the transition activity in linear digital signal processing architectures. The technique extends previous proposed approaches to handle a generalized class of correlated and non-necessary Gaussian data distributions. Using the derived models, an estimation technique is proposed and evaluated for practical examples. Bit level simulations results show the adequate accuracy of the proposed approach.