Oral-History:Ron Schafer

About Ron Schafer

Ron Schafer

Ron Schafer was born in Tecumseh, Nebraska on 17 February 1938. The son of second generation German immigrants, Schafer’s early interest and aptitude for math and science earned him a scholarship at Doane Collage in Crete, Nebraska. After earning a bachelor’s and a master’s degree at the University of Nebraska in Lincoln, Schafer went on to complete a Ph.D. at MIT. During his time at MIT Schafer was one of the first engineers to work with digital signal processing and his theories expanded the concept and applications of the cepstrum beyond the original work done by Tukey and Bogert. Schafer was later recruited by Bell Laboratories where he worked on the application of homomorphic signal processing ideas to the problem of Format Analysis in speech. Schafer's Digital Signal Processing collaboration with Alan Oppenheim was published in 1975. Schafer became a full professor at Georgia Tech in 1974 and later co-authored a textbook on digital processing with Larry Rabiner.

Interviews of Schafer's colleagues provide further information on research at Bell Labs and Georgia Tech, as well as on early digital signal processing. See Alan Oppenheim Oral History, Wolfgang Mecklenbräuker Oral History, Russel Mersereau Oral History, Lawrence Rabiner Oral History, and Manfred Schroeder Oral History.

About the Interview

RON SCHAFER: An Interview Conducted by Richard Nebeker, Center for the History of Electrical Engineering, 14 May 1998

Interview #343 for the Center for the History of Electrical Engineering, The Institute of Electrical and Electronics Engineering, Inc.

Copyright Statement

This manuscript is being made available for research purposes only. All literary rights in the manuscript, including the right to publish, are reserved to the IEEE History Center. No part of the manuscript may be quoted for publication without the written permission of the Director of IEEE History Center.

Request for permission to quote for publication should be addressed to the IEEE History Center Oral History Program, 39 Union Street, New Brunswick, NJ 08901-8538 USA. It should include identification of the specific passages to be quoted, anticipated use of the passages, and identification of the user.

It is recommended that this oral history be cited as follows:

Ron Schafer, an oral history conducted in 1998 by Frederik Nebeker, IEEE History Center, New Brunswick, NJ, USA.

Interview

INTERVIEW: Ron Schafer

INTERVIEWER: Frederik Nebeker

DATE: 14 May 1998

Childhood and family

Nebeker:

You were born on the 17th of February, 1938 in Nebraska?

Schafer:

Yes.

Nebeker:

Can you tell me a little about your family?

Schafer:

My parents are William and Esther Schafer. They are both, I guess you would call them, second generation German immigrants. My father's father and mother both came from Germany and settled on a farm in Nebraska. My mother's mother and father were both born in Germany as well. They lived in a farming community that was mainly German settlers.

Nebeker:

Was it Tecumseh?

Schafer:

Tecumseh is the town that I was born in. The town that was close to where they lived was called-- the Americanization of it was "Steener." Steinauer is the name of the little town.

Nebeker:

Was your father a farmer?

Schafer:

Well, he grew up on a farm. My mother and father were married about 1933. It was during the Depression. They tried to farm and weren't able to do that. My father then had various jobs. He worked on the railroad, called himself a gandydancer. He worked during the Second World War in a munitions plant in Hastings, Nebraska, building storage places for bombs or shells. Then, we moved from the little town of Steinauer to Tecumseh when I was in kindergarten, about five years old. We lived there all of my life. My parents still live there. My father is 91 and my mother is 90 now. My father was basically a truck driver the rest of his working life. He worked first for C. A. Swanson and Sons Company, an Omaha poultry processing company. Later, they were bought by Campbell Soup Company. He worked for them for 25 years or more as a truck driver.

Education

Undergraduate and master's

Nebeker:

So, you went through school in Tecumseh?

Schafer:

Yes, high school and grade school.

Nebeker:

How did you get the idea of going into engineering?

Schafer:

I've thought about that. It was mainly because I like math and I like science. I had a science teacher who encouraged me in that direction. His name was Robert Cook. My math teacher was Sue Boyden. They encouraged me, and I was always a good student. When I graduated from high school in 1956, my parents didn't have much money, so any kind of scholarship that I could get was attractive. I took a test for a scholarship at Doane College. Doane College is in Crete, Nebraska. It's a small liberal arts college. It was, and I guess still is, affiliated with the Congregational Church. It was already a pretty old school and it had about 350 students. But, they offered me a scholarship of $1000.00 for four years of education which allowed me to be able to go to college. Although they didn't have an engineering program, I kind of had that in my mind. They had what they called a three-two program with Columbia University. I thought I would participate in that. When I arrived as a freshman there, it was also coincident with the arrival of a retired professor from the Naval Academy whose name was Levi T. Wilson. He was already probably 75 years old at that time. He had retired from the Naval Academy because he had to, and then he went to Emory University in Atlanta. He taught there, and finally ended up in this little back water of Nebraska, because he just couldn't quit teaching. So, he was the math and science department of Doane College. He was a wonderful man and influenced me, to be a teacher. He taught all of the math courses and all of the physics courses. I took all of them. At the time, they had this notion of a pre-engineering program, so they provided what was one of the key courses in the late '50s: mechanical drawing for engineers. They felt like they had to provide that, so that the students could go to Columbia or a University in Nebraska, as in my case, and not be lacking that skill. So, they hired a civil engineer from Lincoln to come down and teach a course at night in mechanical drawing. I took that course from him and he was looking for someone for his staff. He was just a lone guy who did construction project for small towns in Nebraska. He hired students to do the work. So he hired me for the summer, in 1958, I guess it was.

Nebeker:

In Lincoln?

Schafer:

Yes. He had his office in his basement in Lincoln. At that time I met another well-known electrical engineer whose name is Don Cox, who is now a professor at Stanford University. He worked at Bell Labs for many years. He is an expert in mobile communications. Don was a year ahead of me and he was at the University of Nebraska. I had a high school friend who was going to the University of Nebraska, and she wanted a ride back home. I had a car. Lincoln is 60 miles from Tecumseh and I gave her a ride. It turns out that she had a friend, a sorority sister whose name was Dorothy Hall, and she had the notion that maybe Dorothy and I were a match, so she likes to take credit for matching us up. Dorothy is now my wife. The Tecumseh girl's name was Carolyn Lang. In any case, that is how the connection got made. Instead of going to Columbia University, it looked more appealing to me to go to the University of Nebraska. I continued to work for the civil engineer through all of my college years at Nebraska.

Nebeker:

When did you transfer?

Schafer:

After my third year at Doane in 1959. Then I graduated with a bachelor's degree from Nebraska in '61, and then at the end of the first semester in '61 stayed there for a master's degree. That is where I really got my first taste of teaching. I was a teaching assistant. It got so that I really thought that was what I would like to do. So, I got my master's degree there.

Nebeker:

In electrical engineering?

Schafer:

In electrical engineering. In the meantime I had married my wife, Dorothy, in 1960, and then she graduated in '60 and became a public school teacher, and supported us while I went to school. I stayed on one more year at the university in 1962-1963 as an instructor of electrical engineering.

Ph.D. studies at MIT; digital signal processing

Schafer:

Then in '63, I went to MIT for graduate work. My father-in-law advised me that I should go to the best school that I could get into, so I applied to a lot of places and fortunately was accepted at MIT. I went there and was there for four-and-a-half years completing my Ph.D. degree.

Nebeker:

The field of electrical engineering has attracted a lot of mid-westerners. Have you noticed that?

Schafer:

No. I know a lot of easterners. My friend Larry Rabiner is a New Yorker through and through. So is Al Oppenheim. So, I don't know. It would be an interesting statistical study.

Nebeker:

I heard once that most Silicon Valley entrepreneurs were mid-westerners.

Schafer:

I would hesitate to speculate on any correlation there.

Nebeker:

Ok. So you went to MIT?

Schafer:

One thing that might be true is that many midwesterners left the mid-west to go where things like digital signal processing were being born. Those places were on the West Coast and the East Coast. So, maybe, they kind of stand out as having come from someplace else. I think that it is still true that the opportunities for people in engineering are not in the midwest. They never have been. Maybe there is more opportunity now, than there was in the past. Certainly, that is one of the things that I know that people in the state of Nebraska look at. They say, "Here we have our engineering school. Where do all of our graduates go? Not here." So, you have to take an attitude that the state provided me with a springboard to a good life and a good career, but it didn't do much for the state.

Nebeker:

What did you study at MIT?

Schafer:

I went there to be whatever MIT thought was a good thing to be.

Nebeker:

You decided after the master's that you wanted to do a particular type of electrical engineering?

Schafer:

I wanted to be a college teacher. I knew that you had to get a Ph.D. and develop some ability to contribute to knowledge, and so forth. I went there having an interest in the general areas of circuits, signals, and communication. I didn't have any concept of what I wanted to do because, for one thing, digital signal processing really didn't exist as a field. If you heard Ben Gold's talk, he really taught the first course in digital signal processing. I don't remember what year he did that course, but I didn't take it. The courses that were still prominent, but sort of dying, were courses that had been started by Ernst Guillemin in circuits, network synthesis, and linear systems. I started out as a teaching assistant for Amar Bose, in a course on circuits. It was a wonderful course. Bose was an amazing lecturer. They had money from the Ford Foundation to create this amazing course that every sophomore student took at MIT. There was Amar Bose giving lectures, and Ken Stevens writing notes, and a staff of literally dozens of people. Junior faculty and assistant professors were running recitation sections and there was a cadre of graduate students, each of whom had about 30 students, that they were responsible for, which included grading their homework and giving tutorial sessions. I mean, it was a huge thing. Only MIT and the Ford Foundation could afford such an experiment. MIT continues to use that model, in many cases, for their undergraduate courses. I learned a tremendous amount from just being a teacher there. Then I became involved in a course that Sam Mason taught. He was an amazing character. I was a teaching assistant in his course a couple of times. Then I became involved with Bill Siebert's course on signals and systems. By that time, I had done well enough as a teaching assistant, so that they promoted me to instructor. This promotion meant I was qualified to teach a recitation section. So I began to do that. I got too much into teaching and not enough into research, because that is what you are supposed to be doing as a Ph.D. student.

Nebeker:

It sounds like you were very much interested in teaching?

Schafer:

I was. I loved it. I probably put too much effort into it. MIT had a qualifying exam that they required of every doctoral student. I had difficulty passing that exam because the courses that I had at Nebraska were totally orthogonal to MIT's undergraduate curriculum. They had nine core courses that spanned the whole gamut of things, from circuits to electromagnetic and transmission lines and machines. My knowledge did not match up well with their curriculum. So I had to take a lot of courses. I had never heard of quantum mechanics. So, I had to take an undergraduate course in that. I failed the exam the first time, and barely squeaked through the second time. By that time, I had made contact with Tom Stockham. I had taken his course the first semester that I was at MIT, which was an old Guillemin course. It was electrical engineering 655 and it was linear systems from the Guillemin point of view. Tom was an amazing teacher, very lively and energetic and very knowledgeable, so he influenced me a lot. He was about to go to Lincoln Labs, to become a staff member there instead of a professor. He tapped me to teach this graduate course, that he had been teaching. I taught it a couple of times, once in the summer and once during the school year.

Nebeker:

Is this before you got your Ph.D.?

Schafer:

Yes. Tom was really the key to connecting me to Al Oppenheim who became my thesis advisor. Al was a young assistant professor at the time. Al had been very much influenced by Ben Gold. He had gotten interested in digital signal processing, and was starting out his research career at MIT as a professor, and was looking for a student to try out some of the ideas that grew out of his thesis. Tom was a very good friend of Al's, and he connected me with Al. I had gone to Tom and said, "I would like to work with you and be your student." He said, "That would be great, but I'm not going to be here." So he made that connection. Al and I developed some ideas in digital signal processing. We didn't call it that at the time, but that is where I got started in the field of digital signal processing.

Computers and Fourier-transform calculations

Nebeker:

Was it the personal connection or was it the subject matter of Tom's work that attracted you?

Schafer:

At that point I was very naïve. I didn't know what a research problem was. I was good at teaching subjects that were well understood and developed. It was Tom's ideas that interested me. He had been using computers to process sound. One of the key experiments in the development of the Bose loud speaker was to take sound and music and process it by measuring the impulse response of a room. It was Amar Bose's living room that was measured, by putting spark gaps in the room and firing off a spark which created a kind of doublet wave form that excited the room. Then, you recorded the derivative of the impulse response of the room. The music was sampled, and by digital convolution the impulse response was convolved with the music. Then, you listened to it through headphones, and this is what it would sound like in that living room. These were controlled experiments to try and develop a loud speaker that would have faithful reproduction of sound. There were no fast convolution techniques. There were no fast computers. There was a very rudimentary computer called a TX-0, that had been a research machine. I think it was maybe the first transistorized computer. It was a precursor of the DEC PDP computers. At that time there was the TX-0 and a PDP-1 computer, that were available in the electronics research lab. I gained access to those machines to start to try to apply some new ideas in digital signal processing. When you are doing it, you don't know that you are doing anything special. Thinking back on it, of all the people in the world, I was one of the handful of people who had access to those kind of tools.

Nebeker:

What exactly were the problems you looked at?

Schafer:

At the time, I was starting my thesis work, the Cooley-Tukey paper came out. So it became known that you could do Fourier-transform calculations in a reasonable amount of time.

Nebeker:

Who noticed this?

Schafer:

People like Tom Stockham and Charlie Rader and they of course started spreading the news. Al Oppenheim had done a very theoretical thesis under Amar Bose which was related. In his thesis, he observed that linear systems, which are defined by the principle of superposition, could be generalized. Signals combined by other operations than addition, could also display where the output of the system was. So, for example, if two signals were convolved, convolution obeys many of the properties of addition. It's commutative, and associative. A homomorphic system, as he called it, for convolution would be one where if the input consisted of two convolved signals, the output would consist of two convolved signals. In that case, the system operated independently on the two signals to produce the output. Convolution becomes multiplication. In the Fourier-transform domain, if you take the log of the Fourier transform, you get the sum of a product. The log of the product is the sum of the logarithms. So you break the signal into a sum of two components, which could then be processed by a linear system. Then going back through an exponential, you end up with the product again in the Fourier domain or the convolution in the time domain. We began to look at how we could use and implement these ideas, given the fact that you could now compute the Fourier transform. That was the key. The tools that we had available were very rudimentary computers by today's standards, but they were state of the art at the time. They had a graphics display which was just a device that plotted dots on the screen. It had an A to D and a D to A converter, and a disk. So, we began to play around with these ideas, and that is what ultimately became my thesis idea. The concept of the cepstrum, which you may have heard of, came out of that. It had sort of been proposed independently by Tukey and Bogert at Bell laboratories, where they had observed that using these new Fourier computational techniques, you could measure a power spectrum and compute the logarithm of it and thereby decompose convolutional type processes into a sum.

Nebeker:

So you came up with the concept of the cepstrum?

Schafer:

The first published paper on the cepstrum was by Cooley, Bogert, Healy, and Tukey. It was recognized by Oppenheim that what they were doing was a homomorphic system, which was something that he had studied in a kind of abstract, theoretical sense. So, my thesis was to take those two observations and see what we could make of it. So we looked at the problem of echoes and echo removal in signals, using these ideas. The cepstrum idea has gone on to have much more importance in speech recognition for example, where it became a very compact representation of the short-time spectrum of speech. One of my contributions in my thesis was to develop a theory of this, and expand it well beyond the original Tukey and Bogert proposal. Parallel with this, Tom Stockham picked up on the homomorphic idea and developed the same ideas for signals that had been multiplied together. Then later he applied the convolutional theory to the blind deconvolution of Caruso recordings and so forth. These ideas all sort of had their genesis about 1965, when we began to play with them on computers, so I consider myself very fortunate to have been at the right place at the right time.

Nebeker:

You completed your degree, in '68?

Schafer:

I didn't complete it until about March of 1968 and then I went to Bell Laboratories.

Bell Laboratories

Speech processing; Formant Analysis

Nebeker:

What did you do there?

Schafer:

I should say that when I finished my Ph.D. degree or saw that I was going to finish, I thought, "Well, why did I do this?" I wanted to go into college teaching, so I started interviewing for college teaching jobs. I had a couple of offers, but then I began to realize that I had really only begun to learn to do research and that that was a major expectation of college professors. I thought I would be better off to get some more experience in a research lab. Jim Flanagan at Bell Laboratories was heading a department of speech processing, Acoustic Research Department, is what it was called. He became interested in me, and hired me to go there and do research on digital speech processing. There was a lot of work going on there on using computers to simulate speech processing systems. I had some background that was relevant. I was hired at Bell Labs to work with Jim Flanagan. By that time, Larry Rabiner had done a thesis at MIT, under Ken Stevens, on speech synthesis. Larry already had strong contacts with Jim Flanagan's department through the MIT co-op program. Larry had been there already for a number of years as a co-op student, and had just joined, the previous year, as a full time member of the technical staff. He was the first person that I met, coming in. We hit it off very well, and started collaborating on the application of some of the ideas that I had worked on in my thesis to speech processing.

Nebeker:

What specifically were you working on at Bell Labs?

Schafer:

The famous Arden House Workshops were started in 1968. Shortly after I started at Bell Labs, there was an Arden House Workshop and I presented some of my thesis work there.

Nebeker:

You went to the first one?

Schafer:

Yes. I was fortunate enough to go to the first Arden House Workshop and give a talk on the thesis work that I had done. It was an example of an application that was made possible by the FFT.

Filter design; voice output applications

Schafer:

Then, Larry and I became more interested in things like filter design and we did quite a bit of work on that together. We also, under Jim Flanagan's influence, began to explore digital synthesis of voice and digital representation, so that speech signals could be a mechanism for storing speech in a computer and used as a computer output device. I sometimes have to chuckle when I sit on the phone and go through some tedious voice menu, and I wonder if it was such a good idea. We then started to demonstrate some application of voice output from machines.

Nebeker:

What were the first applications?

Schafer:

It was a curious application. One of the things that we did was to use our machine to synthesize a voice recording of a wiring list for wiring up telephone apparatus.

Nebeker:

I have heard of that from Manfred Schroeder.

Schafer:

I mean it was one of those applications that had a very limited lifetime because there were obviously better ways of doing it, if you could automate wiring. With a wiring machine, why have a human being sitting there listening to a tape-recording saying, "A 32, B 76," you know making connection between those two points. It was something that seemed to have promise at the time. It's totally nonsensical now.

Nebeker:

The reason that the speech synthesis was useful there is that there were so many different wiring patterns that a person couldn't just record?

Schafer:

Right. It was basically reading a list of connections. So you could have a printed list that said B 47 to A 32, but the person was sitting there with this apparatus in front of them, and a wiring gun, then they had to put the wire into the gun and so forth. If you looked away you would lose your place.

Nebeker:

Well, what I'm wondering is why not just tape record somebody saying that?

Schafer:

You could have a person sit there and read the list, but if the list was in computer form in the first place, then you could automatically have the machine generate this spoken list. You remove the person from the operation. It wasn't a great innovation, but it was an example of an application of voice output from computers. So, we did that. We had a great amount of fun. We had very good facilities, state of the art facilities. At that time, Jim Flanagan started to put together mini-computers that could be used for laboratory experimentation with speech processing ideas.

Acoustics Research Group; research division work environment

Nebeker:

How large was Flanagan's acoustics research group?

Schafer:

That group was reasonably large. When I joined there were maybe 15 people. Let me see if I can think of some of the people. Larry, Joe Hall, Aaron Rosenberg, Mohan Sondhi, Cecil Coker, Jim West, and Gerhard Sessler. There were a couple of departments. Jim had a department. Peter Denesh had a department. They were all under Manfred Schroeder at the time. Manfred was the director. Bishnu Atal was in the other department, not in Flanagan's department at the time. I guess there were around ten people in Jim's department.

Nebeker:

Was it a pretty exciting atmosphere?

Schafer:

Yes, wonderful.

Nebeker:

How long were you there?

Schafer:

I went there with the idea that I was going to stay for two years, and get this broadening of research experience, and then go to a university. I was married in '60, and after we got to MIT and got settled, we had two children. By the time that I moved to Bell Labs, I guess my oldest son was three years old, almost four. Then we got to New Jersey, and we lived in New Providence, a very pleasant place to live and close to the lab. We had a very comfortable life. We had another child, a daughter. There didn't seem to be a whole lot of reason to go looking for greener pastures, because we were very happy. So, I stayed for six and a half years. I was in the research division of Bell Labs. The research division of Bell Labs at that time prided itself on the fact that they were more ivory tower than any university. If you wanted to do research, that was the place to be. If you could get yourself there. I was there, so why leave? The output of our research was papers. I published a lot of papers, mostly with people in the group, Larry, Jim and other people there. By the time six and a half years had rolled around, I had established myself with a pretty good record in publication and in digital signal processing.

Collaborative publication with Oppenheim

Nebeker:

Didn't you publish a book with Al during that period?

Schafer:

I published a book with Al Oppenheim. He began to teach a course in digital signal processing after I got my Ph.D. degree. He had never taught one while I was there. So he taught the first course, the regularly scheduled course in digital signal processing at MIT, and he started developing notes on the subject and said, "Well, why don't you collaborate with me on this book?" I didn't realize what an opportunity it was.

Nebeker:

You were at Bell Labs?

Schafer:

I was at Bell Labs at that time, because he didn't start this really until 1969 or 1970. We started working on this book on the side. In the evenings, I would work on the book at the kitchen table.

Nebeker:

How did you divide the labor on that book?

Schafer:

We did that book very much the way that we have always worked together, and that is, that one of us would do a first draft of a chapter.

[End of tape one, side a]

Schafer:

One of us would write the first draft and the other would then take it and put their imprint on it. The first author would then come along, and see what the second one had done, so every chapter went through three drafts. The project stretched out over a number of years. When we were coming down to the end, Al, being a university professor got a sabbatical to go to Grenoble, France, to finish the book. We had a mechanism for communicating. I had my responsibility and he had his. I was always behind on my responsibilities. I always feel like I can do more than I can actually do, and then I have a flurry of activity at the end, and somehow get it done. This was long before the days of e-mail, so our mechanism for communicating was a cassette recording. He would send me a cassette keeping me up to date on what was going on and enclose an up-to-date manuscript. I would respond by flipping the tape over and recording my response.

Nebeker:

You would tape your responses rather than writing them out?

Schafer:

Right. We still reminisce about these tapes. He got very annoyed at me, because I was way behind. He sent me this tape where he said that I "wasn't upholding my responsibilities." He was really wondering whether he had made the right decision to collaborate with me. I sent him back a tape that said, he just "didn't understand that I was doing something else besides writing this damn book."

Nebeker:

He had a sabbatical?

Schafer:

Yes, he had a sabbatical and here I was trying to be a researcher, and a family man, and doing all of this stuff. But anyway, we got it ironed out. We wrote the book and got it published. It became very successful.

Nebeker:

It is one of the defining books for the field.

Schafer:

Later, much later, when I turned 50, Al had kept one of these tapes. The tape always on the one side said, "AVO arrow RWS." Meaning it was his communication to me. On the back was "RWS arrow AVO." He took the tape to a framer, and had one of these very deep frame-box type things made, and the tape is mounted in there. He gave it to me for my 50th birthday. So, I still have that on my wall. I haven't taken it apart to play the tape, though. I think I will just keep that in my memory.

Nebeker:

So, it was a struggle to get that finished, but you must have done a good job on it.

Schafer:

It turned out well. Ben Gold and Charlie had written the first book on digital signal processing, and Al and Tom Stockham had contributed to that. They had written it more or less as a research monograph to summarize what they had learned on the subject. It wasn't a textbook in the normal sense. It had no home work problems. So, that is what we provided. We provided a book that people could teach courses out of.

Georgia Tech

Schafer:

The book came out in '75, and I got a call from Aubrey Bush at Georgia Tech. Aubrey had been a student at MIT when I was there.

Nebeker:

Any relation to Vannevar?

Schafer:

No. Aubrey had been a student with Y. W. Lee who was a protégé of Norbert Wiener, and part of that whole line of people who were influenced by those ideas that were percolating at MIT during the '40s and '50s. At any rate, Aubrey had been on leave when he was at MIT from Georgia Tech, where he had been a professor. He received his Ph.D. from MIT and went back to Georgia Tech. They had received some money from an organization called the Franklin Foundation, the John and Mary Franklin Foundation. The John and Mary Franklin Foundation was set up by John Franklin who had created a company called the Audichron Corporation. The Audichron Corporation provided time and message machines for the telephone system. It was very much of a Kluge. It was this big rotating drum that had messages stored at various locations on the drum and then read heads. You would piece together the message by switching between those heads. They realized that digital stuff was coming on, and that their mechanical monsters would soon be replaced by digital stuff. But what happened was that the president of the company was not John Franklin any longer, but a person named John McCarty, who was a Georgia Tech graduate. John McCarty was killed in an airplane accident, and in his memory they created a chair at Georgia Tech. They were looking for someone with interests in digital signal processing, sound, voice, and so forth to be the chairholder. So he called me and asked if I would be interested in applying for that job. I thought about it. I talked to Jim Flanagan about it, talked to my wife about it, and decided, "Well, you know, now is your chance to either put up or shut up on this teaching business." So I arranged with Jim Flanagan to take a two year leave of absence from Bell Labs to go try this out and see if I could deal with it, see if I could deal with living in the south, because I hadn't heard too much good about the south. So we went there, and of course I didn't come back.

Nebeker:

What year was that?

Schafer:

September of 1974.

Nebeker:

I take it that you did like it there?

Schafer:

Yes. It turned out to be a very good move for me. They were willing to take me into this chair position as a full professor. I leaped over all of the normal hurdles that people have to go through to become a professor at the university. That was on the basis that I had had a good career at Bell Labs, and had established the kind of credentials that university people look for. Georgia Tech was not the place it is today. It was what I would consider a good regional engineering school at the time. It probably had quite a bit of name recognition mainly due to its fight song and its football team. But it wasn't a real power-house university. There was a new president there named Joe Pettit, who probably is on your list of important people in electrical engineering, who had grown up at Stanford and come under the influence of Terman and become Dean of Engineering at Stanford. They hired him at Georgia Tech, to be the president. He set about to change Georgia Tech in a dramatic way. He spent 15 years or 17 years at Georgia Tech and really transformed the place. He did it by bringing people like me with the interest in research and new ideas into the place, by encouraging people to go outside, get money for research, and build up research programs. He had a very effective vice president for research, named Tom Stelson, who was very innovative and very entrepreneurial in using what money he had available, to really push the faculty of Georgia Tech in the direction of doing research, and coupling that into the educational programs. So I came in at just the right time.

Collaborative publication with Rabiner

Nebeker:

I wanted to ask you about another very influential book of the '70s, the one that you did with Larry Rabiner.

Schafer:

Larry and I had done so much work together that we thought alike. Just about the time that I was getting the itch to go and try this new thing out, we got started on a book, summarizing what we thought was the state of digital speech processing. I left Bell Labs, but we continued to work on it. He remained at Bell Labs and I began to teach courses in speech processing at Georgia Tech. We created a text book in the digital processing of speech signals.

Nebeker:

Did you write this book in the same way that you wrote the book with Alan?

Schafer:

Yes, we both worked on every chapter. I would write the first draft and he would take it over and vice versa. It turns out to be a very good way to do it. I think when you write a book, and one person writes one set of chapters, and another writes another set of chapters, it looks like it.

Nebeker:

Yes, you very often can see that.

Schafer:

By the time you have gone through the book two or three times, all of the quirks more or less have been massaged out, or averaged out.

Nebeker:

Did you find in writing your books that the process of writing changed your thinking about things?

Schafer:

Yes. When you sit down to write about something, you realize how much you don't understand about it. In many instances, where we saw some topic that we thought we understood, we didn't. Perhaps, we generated some new thinking in a given area or a notion that some more research needed to be done there.

Publication and research at Bell Labs

Nebeker:

Was it still the case at Bell Labs that people in the research department were sort of encouraged to write "the" book on some subject?

Schafer:

Yes, publication and that kind of impact was very much appreciated at Bell Labs in the '70s. You know it's curious, Jim Flanagan was a very creative researcher and a very good research manager. He was very much a Bell man, a Bell Labs man, a Bell and AT&T man. He wanted the research to have an impact on the Bell system. But, it was a very lethargic system. You had Western Electric over here making things. You had the development department at Bell Labs that was supposed to take the research ideas, and make them into something. If you had an idea that you thought could have an impact, you had to convince a lot of people to make it happen. I remember one time, we had some results that N.S. Jayant had developed. They were simple ideas, Delta modulation ideas, but they allowed very easy encoding of speech into digital form, and Jim wanted to get a chip made that would demonstrate this. He could never get it done. He was so frustrated. There were many ideas that were being developed. I can still hear him say, "If they would just unleash us." He was referring to the FCC. "If they would just unleash us, we could do so many things." So, I guess finally they unleashed AT&T, and you know what has happened since then. It has literally transformed things. Now there is a much closer coupling of research and product.

Nebeker:

One of the traditional strong points at Bell Labs was the support given to making new devices. They had staff with expertise and all the necessary materials.

Schafer:

I'm sure it happened many times. I just remember that we had a very hard time making an impact with the things that we were doing. Things just didn't move that quickly. It was still during the time when everybody was predicting digital representations. Digital processing and digital transmission were going to revolutionize telecommunications. But, it wasn't doing it in the '70s. Then, of course, things changed and it happened very quickly.

Research at Georgia Tech

Nebeker:

Can you summarize your research areas since you left Bell Labs?

Schafer:

Well, at Georgia Tech I worked with a number of colleagues. Tom Barnwell and I continue to work on speech processing. Russ Mersereau joined our group about a year after I went to Georgia Tech. Tom Barnwell had already been there, and he had set up a laboratory for speech processing. So, I benefited from his efforts coming in. Russ and I did some work together on iterative signal restoration. I have continued to have an interest in speech and audio processing. I had a very outstanding student named Petros Maragos, who is now a professor at Georgia Tech. He is on leave this year from Georgia Tech in Greece. Petros got me interested in the subject of nonlinear signal processing, particularly something called mathematical morphology. So, he and I did quite a bit of work together in that field, and I have continued to have an interest in that. So my interests, after Bell Labs, have broadened.

Graduate students

Nebeker:

Have you had many graduate students?

Schafer:

I can't remember the total number. It's not a huge number by some academic standards. It's probably about a dozen or so Ph.D. students over the years. Some of whom have been very successful. Petros. Aggelos Katsaggelos is a professor at Northwestern University. Steve Kay is a professor at the University of Rhode Island. I have a number of students in industry. Craig Richardson is president of a small company in Atlanta. I have always enjoyed working with the doctoral students, because they teach me more than I teach them.

Signal Processing Society, IEEE

Nebeker:

I wonder if we could turn now to the Signal Processing Society.

Schafer:

Sure.

Nebeker:

That is something that you have given a lot of time to over the years. Can you remember the year that you were president of the society?

Schafer:

I think it was '78 and '79, if I'm not mistaken. I can say some positive things about that.

Nebeker:

I would like to hear the negative too.

Schafer:

Well, the negatives are that that was a very troubled time in my life because we had a son who got cancer in his 13th year, which was 1978. So, '79 was a very, very hard year. He died at the end of that year. So, I would probably have to say that I was probably not the most effective president of the Signal Processing Society, particularly in that year. It was a time when there was kind of recognition that Signal Processing was becoming a world-wide topic. There was a lot of ferment, a lot of pressure from people in Europe saying, "Why isn't IEEE doing more over here?" We had some very energetic young people. Many of them were educated in the United States and became interested in creating workshops and pushed to have an international meeting in Europe. A lot of that was going on during the period when I was president. A lot of it came to fruition later with the first meeting of the society outside the United States. I think it took place in Paris, and that must have been in 1982, or was that '84?

Nebeker:

I was thinking '84.

Schafer:

I was thinking '82.

Nebeker:'

82.

Schafer:

We had the meeting in Atlanta in 1981 and I was chairman.

Nebeker:

Was that a good experience?

Schafer:

Yes. I had help from my colleagues at Georgia Tech. By that time, we had other people who had joined us, such as Monty Hayes and Mark Clements, Russ Mersereau, and Tom Barnwell. Everybody played a role in putting on that meeting. It was small by today's standards. I think we had about 900 hundred people at that meeting. We held it in a fairly small hotel. That was a time when the focus had started to shift. I remember that we had a session at our meeting which was entitled, "VLSI: the real hope for signal processing." It was kind of prophetic, but I think the writing was on the wall.

Nebeker:

Was this '81?

Schafer:

Yes. Other people said, "Well, maybe signal processing was the real hope for VLSI." Meaning what in the heck are you going to do with this technology. It is kind of a symbiotic relationship.

Nebeker:

When did you join the society?

Schafer:

Well, I have been a member of IEEE for 35 years, my membership card says. I was a student member at Nebraska, and I may have let it lapse. But, I think I joined up again when I was at MIT.

Nebeker:

Was the audio group or audio and electro-acoustics group taking up the subject of digital signal processing at the time?

Schafer:

Yes. I really became active in it when I joined Bell Labs.

Nebeker:

Weren't you part of the group that worked on the standardized terminology?

Schafer:

Yes. I was a member of the Digital Signal Processing Committee, and we used to have three or four meetings a year. They would be at IEEE headquarters in New York. It was kind of an ingrown group, because it was Bell Labs people, MIT, and Lincoln Labs people, and some IBM people. Those organizations supported us. It was part of my job to be involved in these kinds of things. Bell Labs had a very enlightened attitude on that. Just as long as you didn't spend too much time on it. We got a lot out of it, because we had the interaction with people from these other organizations who were telling us about their latest work. We did a number of projects that were very influential. It was a great experience to be involved with that group of people.

Nebeker:

Would you say that the group that worked on audio and electro acoustics really made a difference, through their publications and workshops, on the rapid development of DSP?

Schafer:

Yes. I think so. IEEE has always had a great deal of respect among the members of the Electrical Engineering Society. If it gets on the radar screen of IEEE, it must be important.

Nebeker:

It's also a matter of finding an institutional home for some new technique?

Schafer:

Exactly.

Nebeker:

I mean, it might have been picked up by some other group within IEEE or outside.

Schafer:

Certainly it might have been. I mean the Circuits and Systems Society might have been a reasonable place for this to happen and they came into it a little later.

Nebeker:

Or the Acoustical Society outside of IEEE might have been a place for it to be really developed.

Schafer:

Yes. At the time that all of this was happening, the Acoustical Society took a very cool attitude towards these things. They weren't basic enough. They weren't scientific enough. So, to get a paper into the Acoustical Society journal on speech processing, it had to really not be a techniques paper, or a paper that showed how you could take speech apart. Larry and I published a paper in their journal on our forma-tracking scheme. But, it was very hard to get papers on signal processing aspects of speech into the Acoustical Society. They were more interested in making measurements on phoneme durations, and whether this sound followed that sound, and so on. They weren't at all interested in signal processing.

Nebeker:

If one takes the view that they are the scientific society interested in the science of speech, and not the techniques for processing and so on, then they were entirely justified.

Schafer:

Right. But, any organization is influenced by the people in it. If the people running things, doing the editing, and so on are not interested in something, then its pretty hard to penetrate that. The people who were interested, wanted to do things, and had the energy, found a home in this audio and electro-acoustics society or group. Like any group, the people that step forward to do the work will be able to, ultimately, make things happen. This is particularly true in IEEE. The people who want to spend the time and devote their energies to it can make things happen. This is certainly my view of what happened with the audio and electro-acoustics group. Here you had people like Bill Lange who recognized the importance of these ideas, the signal processing ideas. You had a very good combination of people and institutions working on these questions. You had IBM, a premiere company who had lots of power, lots of resources. They had connections to Arden House and Columbia University. So Bill Lange organized this highly influential workshop. It brought a lot of people together who reinforced each other. But, IBM, Bell Labs, and MIT had a very strong influence. Some people may have resented that at the time. "What about the rest of us out here, you know?" Probably that was a reasonable response, in some sense, but it did get things going.

Nebeker:

I have taken a lot of your time already. Is there anything that I haven't asked about that you want to comment on?

Schafer:

I just feel fortunate that I have been able to participate in all of this, and make a contribution, here and there. I get a lot of gratification out of the books that I have written with Al Oppenheim and Larry. People come up to me and say things like, "Well, thanks for writing that book." It's pretty good for an author to get that kind of response. I have been pretty lucky along the way, to be involved in this at just the right time.

Nebeker:

Well, thank you very much.

Schafer:

Sure.

[End of interview]