Research topic summary for journal papers

After a busy September in Limassol for the ICANN 2009 conference and the draft of MiCRAM II proposal, it is the time to think about the next journal papers. Following topics are summarised:

• Sound localisation with sustained regular cell and onset cell.
  

• The inferior colliculus (IC) network features with four types of cells. classical system: time-invariant system, now: time-variant system. The possible network structure in the IC is written in notes in 2009-09-3 notebook. The main features of IC cells are: 1) one cell has multiple states, 2) some cell has integration feature on the input. such as the rebounce cell. 3) the combination of 1 and 2 drive the cell to response the input depending to time.
  

• Inhibited MSO model. I have done the main code which shows the shifting of spike rate slope. in next step, I should try the frequency-depedent conductance value. to solve the the problem of no shifting in high frequency such as in 500hz.
  

• The feature or functions of IC network for sound recognition. For example, study the learning feature or adaptive feature to the speech recognition. To use IC network to extract feature which was hardly to get by classical methods. To strength the existing features for speech recognition.
• Using memrist to realize the four type of IC cells 
• Visualize the network structure. see the notes in the ICANN2009.
• 
  
  

Acoustics in the News (Fall 2009)

From Fall 2009, The acoustical society of America

• Ear identification scan could one day be a commonplace technique to identify a caller on the telephone, according to a technology note in the 11 April issue of New Scientist. The concept relies on the fact that the ear not only senses sound but also makes noises of its own, albeit at a level only detectable by sensitive microphones. Such otoacoustic emissions (OAEs) emanate from within the cochlea, and they are thought to be produced by the motion of the hair cells within the outer part of the cochlea. Sounds entering the ear cause these hair cells to vibrate, and these vibrations are converted to electrical signals. OAEs can be provoked when a series of clicks is played into the ear. The returning sound emissions comprise signals between 0 and 5 kHz. The clicks could be supplied by a telephone to identify the person at the other end.
• A computer chip modeled on the human ear could be used in universal receivers for radio-frequency signals, a technology note in the 13 June issue of New Scientist tells us. Devices such as cellphones or FM radios are generally tuned to only a narrow frequency band. The new devices, inspired by the network of hair cells in the ear, can pick up a wide range of frequencies. A network of transistors in the device act like hair cells in the ear. Frequencies ranging from 600 MHz to 8 GHz requires no more electrical power than single-frequency receivers.
• 
  

Does Auditory have persistence phenomenon such as in Vision

As we know, the persistence of vision is phenomenon of the eye by which an afterimage is thought to persist for approximately one twenty-fifth of a second on the retina [Wiki]. Film is made based on this phenomenon in order to make people feel the motion.

I am curious whether human auditory system has the similar phenomenon. Persistence in cochlea or in the inferior colliculus. Let me call it persistence in auditory. This can help to explain the "old-plus-new" (P371 in Bregman's Auditory scene analysis) Heuristic. Old stream is persistent for a short while and the difference between present and past is thought new. There is very limited research on it, Bregman mentions kind of pitch persistence in his book (P24) and P750 reference (Persistence of a pitch-segregating echoic memeory. Journal of Experimental Psychology: Huamn Perception and Performance, 2, 531-537. Kubovy, M., and Jordan.)

This may also help to understand the different performance of people in speech recognition . Fluent listener may have long persistence time in his pre-cortex brain which can combine the feature in a longer past into a complex representation of the sound. In contrast, beginner has short persistence time in pre-cortex and it is not enough to build a feature before cortex and it asks the primary cortex to mentally keep persistence which will cost process time. So that beginner need slow speed.

It may also help on the percedence effect (P154, Wang's Computational Auditory Scene Analysis) that directional cues due to the direct sound are given a higher perceptual weighting than those due to reflected sound. This can be explained by using both persistence of auditory and old-plus-new. The direct cues is persistence for a while and it will compare to the reverberation coming later. If this is no new spectrum, the brain will think it is still the old stream rather than a new one so that to ignore the reverberation or to mask it (see the old-plus-new Heuristic).

Google book is a good research tools

I have been used google books for a year. However, I haven't found its benefit until today. I thought it is just a collections of magazine and 3rd-level books.

Today, I tried it for my literature view on "sound reverberation on perception". It does really impressive work. First, I searched the book I am reading in hard copy, "Auditory scene analysis, (Albert S. Bregman)". This book is of partly view in Google, though I and search the key words in the whole text. It helps me to quickly locate the page in the hardcopy.

Great works, google.

By a way, I notice that Bregman haven't done much research on reverberation. More related works can be found in Spatial Hearing by Jens Blauert.

At the beginning of this blog

This blog will act as a log to record progress or hint in my daily academic research. Welcome to put comments and suggestions. My research is related to robotics, vision, auditory, neural networks and artificial intelligence.