Possible reconstruction of auditory signals by the central nervous system (CNS)
Richard A. Altes · The Journal of the Acoustical Society of America · 1978
Tuned auditory neurons and behavioral sensitivity to the envelopes of neuronal responses rather than to phase-locked discharge patterns [W. M. Siebert, Proc. IEEE 58, 723–730 (1970)] suggest that the CNS receives a spectrogram-like representation of acoustic stimuli. How much information about the original input signal u(t) is preserved in such a representation? The spectrogram is the convolution of the signal's time-frequency energy density function eu(t,f) with that of the filter which is used to construct the spectrogram [M. H. Ackroyd, J. Acoust. Soc. Am. 50, 1229–1231 (1971)]. By using two-dimensional deconvolution, eu(t,f) can be obtained from the spectrogram. The signal's auto-ambiguity function Xu(τ,φ) is the two-dimensional Fourier transform of eu(t,f), and u(t)exp(jλ) can be obtained from Xu(τ,φ), where λ is an arbitrary constant [W. M. Siebert, MIT, Q. Prog. Rep., 90–94(1958)]. If the spectrogram model is valid, these observations imply that the CNS can, in principle, completely reconstruct the input signal except for an unknown constant that is added to the phrase of u(t). It follows that u(t) and u(t) exp (jλ0) should be monaurally indistinguishable for any λ0.