A comparison of five methods for computing the power spectrum of a random process using data segmentation

When we compute the power spectrum of a long input vector with limited core memory it is necessary to segment the data. Five alternative methods are possible. This paper presents a number of computed results comparing the resolution, stability, and overall quality of these methods.