Subject: Re: AW: Cochlear nonlinearity & TTS From: Ramdas Kumaresan <kumar@xxxxxxxx> Date: Thu, 18 Jan 2007 17:02:34 -0500 List-Archive:<http://lists.mcgill.ca/scripts/wa.exe?LIST=AUDITORY>It would seem that signal amplitude (how big it is ?) and the signal shape, that is the details, are two different things. As an engineer, if I were designing the cochlea/auditory system, I would extract the amplitude info separately in each frequency region, and scale the signal (amplify it if necessary) to a nominal value, say, to unit amplitude, and then try to represent its shape in the spike patterns. This can be thought of as lossless compression/expansion followed by signal shape representation. In the brain actual decompression need not take place. All that you need is info regarding the bigness of the signal. That is also a feature after all. It would also seem that at the periphery stage there is not enough intelligence in the mechanics to determine what is the “feature” that is important down the line in the system. It would make sense for the periphery to blindly “represent” what ever happens to come by in the form of an input signal. However it might be that it does not have the resources to represent all possible input signals uniquely. But it should try. That is, when you have a huge tone and a little tone that happens to fall in its shadow, clearly the periphery has to make a choice to mask the weak one. However this kind of masking madness must have a method behind it. Is it not true that the periphery always represents any signal input uniquely (that is even a combination of a strong tone with a weak one nearby, that is if you examine all the nerve fiber outputs there will be evidence of the weak tone) but in further processing down the line, the fine details in the auditory fiber outputs are lost and thus masking takes place? RK Quoting Erik Larsen <elarsen@xxxxxxxx>: > From that point of view, the cochlea might be better described as an > acoustic feature extractor rather than a compression system. The latter > would imply that at some point decompression occurs, yielding a > (approximate) replica of the acoustic signal somewhere in the CNS; I > would guess this does not happen (what would be the point?). > > If we agree that the brain is only interested in specific features of > the acoustic signal (those that we perceive as pitch, loudness, > location, timbre, etc), there is no need for a compression/decompression > system. It would be more efficient to extract the features as accurately > as possible without considerations as to whether the operations are > invertible. > > Whether it is possible in principle to reconstruct the acoustic signal > from the nerve output is a different question. You would definitely need > the spikes from all nerve fibers (at least the more, the better), and > you won't be able to get it exactly right (if you don't have a priori > knowledge about the signal). > > Erik > > -- > Erik Larsen, Ph.D. candidate > Speech and Hearing Bioscience and Technology > Harvard-MIT Division of Health Sciences and Technology > Cambridge MA 02139 > http://web.mit.edu/shbt > > David Mountain wrote: > > believe that the available data suggest that the cochlear amplifier only > > has a significant effect over a small region at and basal to the peak of > > the traveling wave. See for example: > > > > Cody AR (1992) Acoustic lesions in the mammalian cochlea: implications for > > the spatial distribution of the 'active process'. Hear Res. 1992 > > Oct;62(2):166-72. PMID: 1429258 > > > > I would rephrase Ramda's question to be the question of whether the > > transformation between sound and auditory nerve firing pattern is > > invertable. I think the answer in the exact sense is that it is not > > invertable. It is my belief that we are dealing with a compression system > > with some loss but one that preserves the features of biological > > relevance. Modern audio compression schemes take advantage of this fact > > and throw away acoustic information that is not perceived or only barely > > detectable by the listener. > > > > > > -------------------------------------------------------------------- > > > > David C. Mountain, Ph.D. > > Professor of Biomedical Engineering > > > > Boston University > > 44 Cummington St. > > Boston, MA 02215 > > > > Email: dcm@xxxxxxxx > > Website: http://www.bu.edu/dbin/bme/faculty/?prof=dcm > > Phone: (617) 353-4343 > > FAX: (617) 353-6766 > > Office: ERB 413 > > On Thu, 18 Jan 2007, Ramdas Kumaresan wrote: > > > >> Navid, Richard and the listees, > >> > >> I have heard a lot of speculation about the cochlear amplifier for many > >> years. One of the questions that I have wondered about > >> as a signal processing engineer for many years, is with all the > >> sophisticated nonlinearities, delays, amplifiers, filters > >> etc that are present in the auditory periphery, how does it "represent" > >> an acoustic signal in the neural spike patterns > >> that emanate from the auditory periphery? (I guess everyone wonders > >> about it.) > >> Is it possible to reconstruct the acoustic signal if you were able to > >> measure/monitor the > >> spike patterns that are put out by all the auditory nerve fibers? What > >> is the reconstruction 'algorithm"? > >> (I know about Egbert deBoer's reconstruction method for a single > >> nerve fiber.) Is'n't the information about the signal > >> distributed across many, many nerve fibers? Should'nt the > >> reconstruction take information from > >> all nerve fibers and fuse them to reconstruct the signal? Just wondering > >> aloud. RK > >> > >> > >> > >> > >> > >> Richard F. Lyon wrote: > >> > >>> At 9:17 AM -0800 1/16/07, Navid Shahnaz wrote: > >>> > >>>> Thank you Reinhart for your clarification. Does the cochlear > >>>> amplifier works on both sides of the excitation pattern peak on the > >>>> BM? or the amplifier operates wore efficiently at a place that is > >>>> just above or toward the apex from the point of disturbance created > >>>> by travelling wave? Operationally this point may be an ideal point as > >>>> it is less likely saturates the amplifier due to sharp slope of the > >>>> travelling wave on the apical side. > >>>> Cheers > >>>> Navid > >>> > >>> Navid, > >>> > >>> Both Monita and Reinhart have given good explanations, but let me add > >>> a bit. > >>> > >>> The way I think of it, the active amplification is active everywhere, > >>> but it competes with the passive loss mechanisms, and is only > >>> significant at low enough levels. The active loss mechanism (damping) > >>> increases rapidly apically when a sine wave travels past a > >>> characteristic place. Because of the active gain, the response to a > >>> sine wave can travel further before it damps out; from the "passive > >>> peak" that Reinhart mentions, the peak response location can be > >>> further apical, up to about a half octave worth of place further, when > >>> the active amplification is significant, to the "active peak". The > >>> "net" amplification is positive (in dB per mm or whatever) before the > >>> response peak, and negative after the response peak, pretty much by > >>> definition of peak. That net includes the active gain, which > >>> saturates, and the passive loss, which doesn't, so it's level dependent. > >>> > >>> In addition to the saturation that reduces the active gain at high > >>> level, there is also efferent control that turns down the gain in > >>> response to afferent response level and possibly other central control > >>> signals. This effect of efferent control of mechanical gain has been > >>> directly demonstrated, but I don't recall exactly who/when/where to > >>> cite right now. > >>> > >>> Dick > >>> > > -- > Erik Larsen > PhD candidate Speech and Hearing Bioscience and Technology > http://web.mit.edu/shbt > > 'The scientist does not study Nature because it is useful; > he studies it because he delights in it, and he delights in it > because it is beautiful. If Nature were not beautiful, it > would not be worth knowing, and if Nature were not worth > knowing, life would not be worth living.' > > -Henri Poincare > >