Изображения страниц
PDF
EPUB

The information handling specialist might find such proposed synthetic image processing equipment, which has been intentionally patterned after biological systems, a novel research tool for the study of image processing techniques and theory.

And finally, the computer designer might find that the current generation of blind and data spoon-fed computers could use visud equipment of its own. This analogy can be stretched further by noting that a number or computer functions, currently handled in the interior of computers might be advantageously separated and incorporated into their sensory input equipment, just as the retina separated from the brain during the biological evolution.

In conclusion we would like to suggest that an organic blending of both synthetic and analytic research approaches promises a rapid application of biological principles to the practice of engineering arts.

The writer would like to acknowledge many helpful discussions with a great number of his colleagues at the RCA Laboratories. He feels especially indebted to Mr. Hook and Dr. Amarel for their constructive criticism.

List of References

1.

2.

3.

4.

5.

6.

E. E. Loebner, "Physics of the Solid State in Electronics and
Tellecommunications" Vol. 4, p. 762 (1960).

L. S. G. Kovas znay, and H. H. Joseph, Proc. IRE 43, 560 (1955).

R. A. Kirsch, L. C. Ray, L. Cohn and G. H. Urban, Proc. EJCC, 221 (1957).

S. H. Unger, Proc. IRE 46, 1744 (1958); 47, 1737 (1959).

J. S. Bomba, Proc. EJCC 18, 218 (1959).

R. A. Kirsch, M. Minsky and U. Neisser, Proc. EJCC 18, 233 (1959).

7. Noam Chomsky, Syntactic Structures, Morton and Company, S. Grovenhage (1957).

[blocks in formation]

9. Stephen Polyak, The Vertebrate Visual System, The University of Chicago Press, (1957).

10.

J. R. Platt, American Scientist, 44, 180 (1956).

J. R. Platt, Symposium on Information Theory in Biology, p. 371, Pergamon
Press 1958.

11.

12.

13.

14.

J. Y. Lettvin, H. R. Maturana, W. S. McCulloch and W. H. Pitts, Proc. IRE, 47, 1940 (1959).

H. K. Hartline, J. Gen. Physical 40, 357 (1957); IRE Trans. Med. El. 6,
84, (1959).

A. Rose, Journal Soc. Mat. Pict. Eng. 47, 273 (1946).

15. V. K. Zworykin and L. E. Flory, Proc. Am. Phil. Soc. 91, 139 (1947).

16.

A. Rose, J. Opt. Soc. Am. 38, 196 (1948); 43, 715 (1953).

17.

0. H. Schade, Jr., J. Opt. Soc. Am. 46, 721 (1956).

[blocks in formation]

20. W. S. McCulloch, Transactions of the Symposium on the Mechanization

21.

22.

23.

24.

25.

of Thought Processes, Teddington (1958).

B. Kazan and F. H. Nicoll, Proc. IRE 43, 1888 (1955).

F. H. Nicoll and A. Sussman, Proc. IRE 48, 1842 (1960).

H. O. Hook, RCA Review 20, (1959).

Final Progress Report, Research on Optoelectronic Computer Components,
Contract No. AF19 (604)-4959, Air Force Cambridge Research Center

Air Research & Development Command.

F. H. Nicoll, RCA Review 19, 77 (1958).

NATURE'S CONTRIBUTION TO CORRELATION PROCESSES

L.A. de Rosa

L.M. Vallese

International Telephone & Telegraph

INTRODUCTION

In recent years considerable developments have taken place in communication theory, based on the utilization of statistical approaches. The fundamental concept consists of devising practical processes for the resolution of messages and noise on the basis of their statistical properties. In this connection, one of the most effective processes is that based on the use of the correlation function. Although from a theoretical point of view such approaches cannot be claimed to be superior to conventional filtering in the frequency domain, in practice they may be more advantageous in certain cases where better engineering approximations can be realized.

One of the outstanding examples of correlation type processing of the information is represented by the human auditory function. It has been suggested that the behavior of the cochlea may be interpreted on the basis of a simple autocorrelation function similar to that found in most man-made correlation detectors. Actually, however, the behavior of the ear presents a number of peculiar aspects which indicate a more complex type of operation. It is the purpose of the following paper to discuss a possible interpretation of the said functioning, which constitutes the basis for the realization of an analog of the system.

PRINCIPAL PROPERTIES OF THE HUMAN AUDITORY SYSTEM

The principal part of the auditory system is represented by the cochlea (Figure 1). This is a spiral passage winding into the base of the skull and provided with a canal for the transfer and processing of the sound waves. The canal is divided along its length by a strong membrane, called the basilar membrane, on the surface of which are located the hair cells that constitute the Corti organ. The two major subdivisions of the cochlear canal, called respectively the scala vestibuli and the scala tympani, are filled with a liquid (called perilymph) and are connected by a small opening at the top of the cochlea, in the apex of the spiral.

The range of operation of the auditory system is limited by the threshold of feeling. The first is an indication of the sensitivity of the ear, and the second is a protective upper limit of response set up to prevent damage. Both characteristics, shown in Figure 2 for a "median" ear, indicate non-uniform response behavior as function of frequency. Of particular interest is the observation that relative pressure variations of the order of 10-9 can be detected.

The response of the ear to relative variations of the sound intensity and to relative variations of sound frequency are rather complex. In Figure 3 is shown the minimum perceivable change in intensity for each frequency and in Figure 4 is shown the minimum perceivable change in frequency at constant intensity. Various attempts have been made to represent these relations analytically. The most interesting aspect of such response is represented by the phenomenon of localization (Figure 5); if a single frequency sound stimulus is considered, the distribution of the sensation along the basilar membrane follows sharply peaked curves, the selectivity of which appears to increase rather than decrease with the increase of the intensity of the sound. On the basis of this phenomenon of localization, it is possible to determine the number of distinguishable frequencies at each intensity level. For example, there are 1600 frequency steps at 60 db intensity level (Figure 6).

Even more complex behavior is found in correspondence of multifrequency sounds. For example, if two sounds of different frequencies are present simultaneously, the threshold of audibility is modified. More specifically, the lower frequency sound tends to reduce the response to the higher frequency sound, and in certain conditions produces complete "masking". In Figure 7 the threshold curve distributions produced in the presence of a 200 cps and of a 1200 cps sound are plotted. Due to the complexity of these phenomena, the present investigation will not take them into consideration and will be limited to the realization of the result of localization. This effect appears to be mainly responsible for the resolution characteristics possessed by the

ear.

THE ANALOG OF THE COCHLEA
(1)

In a previous paper it was noted that the sound pressure stimulus sets up two pressure waves respectively in the scala vestibuli and in the scala tympani, which possess different elastic and dispersive characteristics. As a result, the perturbation in each canal is the superposition of two separate components having different phase velocities. Physically, it is understood that as these two waves travel with different velocities toward the helicotrema, there occur points along the canal at which both components have maxima of pressure value and points where they both have minima. The points of maximum sensitivity of the basilar membrane are those for which a maximum compression occurs on the side of the scala vestibuli and maximum rarefaction on the side of the scala tympanií. In fact, the nerve motivating action is considered to be the liberation of inhibitors, and this would take place more significantly at those points.

From the point of view of the construction of electronic analogs, one has to consider the propagation of two signals, one of which is delayed with respect to the other by an amount which is a function of the distance along the basilar membrane and of the frequency. Assume that a single frequency stimulus is applied. If the two pressure waves are represented respectively with

[ocr errors][merged small][merged small][merged small][merged small]
« ПредыдущаяПродолжить »