THE CORRELATOGRAPH 1185 



to filter out the strong low frequency components from the input. The 

 ''s" sounds show closely spaced stripes indicating a concentration of 

 energy at the top of the band. Some of the consonants do not show much 

 under the conditions of this test. 



It is of interest to compare correlatograms with spectrograms of the 

 same signal. For the single frequency input, the stripes on the correlato- 

 gram would be replaced by a single line on the spectrogram. It would be 

 possible to construct signals such that these patterns are interchanged. 

 For example, a rounded band of noise transmitted over two paths having 

 different times of transmission would have a correlatogram with two 

 stripes corresponding to lag times of zero and the delay difference while 

 the spectrogram would show a periodic array of stripes corresponding 

 to the interference pattern of the two paths. It appears that there may 

 be complementary fields of usefulness for the two kinds of analysis and 

 further study is planned. 



Besides the many individuals previously named as contributing the 

 various phases of the project, I would like to acknowledge the inspiration 

 and guidance of R. K. Potter in initiating and carrying through the 

 program. L. C. Peterson shared equal responsibility with the author in 

 the project and but for his untimely death would have been a co-author 

 of the present paper. 

 November 6, 1952 



REFERENCES 



1. W. Koenig, H. K. Dunn andL. Y. Lacy, The Sound Spectrograph, J. Acoust* 



Soc. Am., 17, pp. 19-29, 1946. 



2. Y. W. Lee, T. P. Cheatham, Jr. and J. B. Wiesner, Applications of Correlation 



Analysis to the Detection of Periodic Signals in Noise, Proc. I.R.E., 38, pp. 

 1165-1171, 1950. 



3. K. N. Stevens, Autocorrelation of Speech Sounds, J. Acoust. Soc. Am., 22, 



pp. 769-771, 1950. 



