NetNews Usenet Archive 1992 #20/NN_1992_20.iso/spool/comp/ai/3423

Internet Message Format | 1992-09-11 | 9.6 KB

Path: sparky!uunet!mcsun!uknet!edcastle!edcogsci!holger
From: [e-mail] (Holger Hank)
Newsgroups: comp.ai
Subject: Artificial Noses (Summary)
Keywords: noses, AI, neural nets, smell
Message-ID: <[e-mail]>
Date: 11 Sep 92 10:48:07 GMT
Organization: Centre for Cognitive Science, Edinburgh, UK
Lines: 283

Sorry, it took me ages to come up with a summary of the replies to my question"
"What about ARTIFICIAL NOSES?"

It's very rough and ready, apologies.

Plenty of suggestions as to where they are and could be used. Quite a bit on actual systems, research, papers.

I'll sum up replies of the first type, and append the most interesting
stuff about who does what where in the field.

More info on artificial noses appreciated. Please email me.




* Airpost security is intersted in Artificial Noses to sniff out drugs,
bombs etc.

* Wine smelling\testing  systems have been build.

* Smoke alarm systems.

* Crime forensics, smell the criminal ?!?

* In France, they are looking for artif noses to replace expert sniffers
in the perfumery industry.

* Artificial noses could be used in food processing, test whether foods
have gone off (Suggest this as a standard for fridges...).



*****

PROJECTS, etc. (original messages)

To: [e-mail]
Subject: neural nose !!!
Sender: [e-mail]
Status: RO

Dear Holger,

   I read with interest your posting on the notice board. In March of
this year, Nick Ryman-Tubb of Neural Technologies, Petersfield, Hampshire
wrote a paper in th"Electronic Engineering" titled "Designing an electronic
wine-taster using neural networks. In this article he developed neural
network hardware which could see, taste and "smell" the wine and could
distingush between white sweet, white dry, red sweet and red dry. Noi[t
an outstanding accomplishment I know. However, he presented all the
circuits we used for the sensors and the network he used. a 3-4-4, not
very large. The electronic dept at Kings Buildings should have the
magazine "Electronic Engineering" and the neural group headed by
Alan Murray should be able to explain the circuits if you are not familiar
with circuits. Hope this is of some help.

  Tony Smith
  Hitachi Central Research Labs
  Tokyo, Japan.

  ps. I did my PhD in neural nets with Alan Murray in '85.

Subject: Re: Artificial Noses
Date: Mon, 3 Aug 1992 12:16:02 +0200
From: Josef Goeppert <[e-mail]>
Original-Sender: [e-mail]
Status: RO


Hello Holger,

    I am working with a new sensory system, which seems
to be simular to natural smell and taste. The aim is not
to study natural sense, but to devellop cheep sensor
system for online test of waste water and gas. Different
gases interact with polimerical membran and change
thickness and optical properties. By help of optical
interference technics, these parameters are meshured.
Recognition of gases and concentration is made by a
self organizing map.

    There are quite a lot of parallels with matural
taste: 1) single sensors are sensible for several gases
in its specific way. 2) Relation between gas and
sprectra is extremely non linear 3) evaluation with
topologie preseving map is desired.

    This project is my final work of cognitive science
studies, and I am interessted to contact someone, working
in this domain to study relation of my system with
natural smell and taste.

    Curious to your answer, best regards,


        Josef Goeppert
                Wilhelm-Schickard-Institut
                7400 University of Tuebingen
                Germany
                tel: (49) 7071 / 295940
                [e-mail]


Subject: Re: Artificial Noses
Status: RO

[e-mail] (Holger Hank):
> Just out of curiosity:
>
> are there any systems around which can *smell* , rather than "see" or "hear"?
>
> Any info about such systems, research, etc  appreciated. also: What
> would be the use of sniffing computers/robots?

At least one water supply company uses an exotic species of fish which
emits electrical impulses (I forget why) in some way proportional to
the concentrations of pollutants and impurities in water.  They are
effectively used as peripherals to a computer system, and are much
more sensitive that fabricated sensors.  Maybe they can be educated
to discriminate between pollutants?  They work in shifts, spending
some of their day in individual cages within the main water flow.
Don't quote me please - this is just a garbled recollection of
something I saw on television a while ago.
Subject: Artificial Noses
Status: RO

I know of a group from Naval Lab (NOSC) in San Diego working on
Garner and Lynch's (sp) Olfactory model. They have implemented
this in VLSI. Need further info.....


raj patil

Organization: University of California at Berkeley
Cc:
Sender: [e-mail]
Status: RO

In article <[e-mail]> you write:
>Just out of curiosity:
>
>are there any systems around which can *smell* , rather than "see" or "hear"?
>
>Any info about such systems, research, etc  appreciated. also: What
>would be the use of sniffing computers/robots?
>
>please email me, I'll post a summary.

What a coincidence:  I just heard on the radio that Sony has created an
artificial smelling system that, according to the radio story, scored
10 out of 10 on discriminating between different perfumes.

-Michael Braverman
[e-mail]

Subject: Re: Artificial Noses
Newsgroups: comp.ai,comp.ai.neural-nets
In-Reply-To: <[e-mail]>
Organization: Engineering Design Research Centre
Cc:
Status: RO

Check out the work of George Dodd at ? the University of Warwick.
He has been working in this field for many years and did, at one time, have an
artificial nose comprised of a small number ?3 of semiconductor gas sensor devices.
He used the analogue levels from these to associate a characteristic signature
with a small number of volatiles.
----
cheers
Iain Donaldson
Sender: [e-mail]
Status: RO


We are building an agricultural harvesting robot which uses smell
as part ofi ts sensory array.
This project is being done at the Agricultural Engineering dept.,
Purdue University, West Lafayette, Indiana, USA in cooperation with
the Institute of Agricultural Engineering, Volcani Center, Israel

The objective is to build a robot which can travel down the field, and
selectively harvest only the ripe fruit, which in our case is melons.
The robot is called
VIP ROMPER: Volcani Institute-Purdue, Robotic Melon Picker

One of the main problems in selective harvesting is the determination
of the ripeness stage of the fruit. We have built an artificial nose
which is mounted on the end effector, and can determine fruit ripeness
by the amount of aromatic volatiles emitted by the fruit
(nose-in-hand as opposed to eye-in-hand).

The prototype executive controler and vision sensors are currently
being evaluated in field studies, and the nose will be incorporated
later this year.

I believe this is the first smelling robot in existence, except for
one experimentally used in the car industry for detecting leaks in
car bodies - filling them with some  detectable gas first.

Our robot is taught to recognize different varieties of melons since
the aromatic composition differs for the different varieties.

  Meny Benady
  Agricultural Engineering Dept.
  Purdue University
  West Lafayette, Indiana 47907
  USA

p.s. If you indeed distribute a summary, I would appreciate if you
could send me one directly, to: [e-mail]


To: [e-mail]
Subject: Re: Artificial Noses
Newsgroups: comp.ai,comp.ai.neural-nets
References: <[e-mail]>
Status: RO

In comp.ai you write:

>Just out of curiosity:

>are there any systems around which can *smell* , rather than "see" or "hear"?

>Any info about such systems, research, etc  appreciated. also: What
>would be the use of sniffing computers/robots?

>please email me, I'll post a summary.

Malcolm Haskard at the U of Adelaide (microelectronics centre) has
built an artificial nose which is Hebbian.

Tomasso.



Subject: Re: Artificial Noses
Status: RO

I know that a guy at:
 Department of Instrumentation and Analytical Sciences (DIAS)
 University of Manchester Institute of Science and Technology (UMIST)
 Manchester
 England
has built an artificial nose. A friend of mine who studied there told me.
But I have been unable to contact him. I suggest you write to the head of
this group. And if you get any information (on this nose or any other)
please forward this information to me.

Sjur J. Vestli


Subject: Artificial Nose
To: [e-mail]
Date: Fri, 7 Aug 92 13:56:28 EST
Sender: [e-mail]
Status: RO


Hello Hank,

Here in OZ, members of our knowledge base systems program have
recently been discussing a collaborative project with the CSIRO
Division of Food Technology (DFP) in the application of neural nets,
machine learning, or statistical classifiers in an artificial nose
project. The Sensory Research Centre belonging to DFP has a project
which they call the "artificial nose", in which they are developing
a multisensor system to detect various volatiles and aromatics given
off by different food types. The aim is to develop a system for use
in food processing to detect when food has gone bad, etc.. The sensor
is intended to be faster and cheaper than other methods, such as gas
chromatography.

regards, Craig Lindley


--------------------------------------------------------------------
Craig Lindley       ( [e-mail] )
CSIRO Division of Information Technology, Sydney, Australia.
PO Box 1599, Macquarie Centre, North Ryde, NSW, Australia, 2113

Aust. p: 02-887-9441
      f: 02-888-7787