Low-Complexity MDL Method for Accurate Source Enumeration

A low-complexity method for source enumeration is proposed in this letter. Given the training data of a desired signal, an array data matrix is partitioned into orthogonal signal and noise components. The noise components are then used to calculate the total description length required to encode the array data. The model with the minimum description length (MDL) is chosen as the best model. Unlike the traditional MDL methods, the proposed method linearly partitions the array data into the cleaner signal and noise components and thereby is more accurate and computationally efficient. Its performance is demonstrated via numerical results.