Document Type

Thesis - Open Access

Award Date


Degree Name

Master of Science (MS)

Department / School

Electrical Engineering

First Advisor

Robert G. Finch


The recent growth of the field of digital signal processing has spurred research into many new methods of data handling. One of the topics that has received some attention during the past few years is the area of median filtering. Median filtering seems to be well suited to eliminating spike noise in communications signals and is also used in image processing to sharpen the outline of an image. The basic terminology used to describe the median filtering process is given below. That section is followed by a brief explanation of the approach to median filtering used in this project. The concepts involved in median. filtering will be explained in greater detail in the literature review. At this point the general characteristics of median filters will be presented. First, it should be made clear what is meant by the term median. To find the median of a set of numbers, the elements of the set are compared and put into order from largest to smallest. The element that occupies the central location in the ordered list is the median. As an example, assume that the median of the set [1' 7' 2' 8, 5] is to be found. Ranking the numbers generates the ordered list [l, 2, 5, 7, 8]. Since the number 5 occupies the central position in the list, it is the median value. After determining the output of the first window, it is necessary to decide where to place the succeeding windows. A common practice is to slide the window over by a fixed number of data points to get the next window. The number of data points skipped over is arbitrary and depends on the application. For this project, the convention used was to slide the window over by one point at a time. In effect, the oldest data point is replaced by the newest data point after each sample. A filter of this type is called a running median filter since the data runs through the filter as fast as it is sampled. Since the characteristics of median filters have already been well documented, this project did not attempt to delve more deeply into the theory of why median filters work. Instead, the focus was to try to determine how well a particular filter design could be made to work and whether using microprocessors for this task would be feasible. The initial goal was to build a filter that could be used to filter a speech signal. It was hoped that the filter would be capable of running at 20 KHz, which would enable it to filter most frequencies in the human vocal range.

Library of Congress Subject Headings

Signal processing -- Digital techniques




Number of Pages



South Dakota State University