• Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 245
  • Last Modified:

neural networking simple questions

Hello
I have just started a research after 14 years of  leaving university. I am researching into neural networking with some environmental applications. I am at the reading and review stage of neural networks in general. I have some easy questions they will help me put things together. They might seem very easy but I am very new to this subject.

the questions are:

The properties of neural networks that make them very suitable for handwriting/image recognition are
The learning process of neural networks and
The pattern recognition process
are there any other advantages that the neural network are suitable for image recognition?

the second question is related to speech understanding gesture computing


If we want to develop a system for speech understanding (not speech recognition) that can
cope with language use in everyday conversations (face-to-face dialogue).
what examples of language use in such a situation that illustrates that we need a gesture recognition system as part of our speech understanding system. This is an imaginary question. I am not sure if such system exists but I need to know your openions about this please.

Jean

 
0
jean11
Asked:
jean11
1 Solution
 
scottjlloydCommented:
Your questions are a bit ambiguous.  Other than the learning method and pattern recognition what else could there be?  I think you have enumerated all the aspects of NNs.  One strength of NN over other types of pattern recognition programs is its robustness.  Assuming the NN is not over trained to the training set the trained network would (should) be able to correctly identify different type of handwriting.  By this I mean if you train the network using right handed script the network should identify left handed penmanship without retraining.  
Also if errors are encountered with identifications the specific example that was misidentified can be added to the training set and then incorporated into the weights thus "learning".  I have done years of research and written several articles using NN including different training methods and other variations.  So if you can be more specific in your questions I can provide a better answer.

You may want to look at: http://www.cs.stir.ac.uk/~lss/NNIntro/InvSlides.html.  Professor Smith has an excellent primer concerning NNs that is written in a straight forward manner and very easy to understand.

Your second question, more theoretical in nature could probably be answered by looking into the current research on non-verbal communication.  This research details gestures, facial expressions and some variations on this research also include tone as a form of communication that is not necessarily conveyed by just the word choice.  The state of the art for NN and speech recognition is still not to the point where speech is easily recognized.  Individual training must be extensive to recognize speech from a specific individual.  Dragon Natural Speak is the leading computer application of speech recognition and you may want to visit their website for additional information on speech recognition.  Since their program is proprietary it is unknown (but generally assumed) if NNs are used in their software.

Scott J. Lloyd, Ph.D.
0

Featured Post

Concerto's Cloud Advisory Services

Want to avoid the missteps to gaining all the benefits of the cloud? Learn more about the different assessment options from our Cloud Advisory team.

Tackle projects and never again get stuck behind a technical roadblock.
Join Now