So i hope you are ready for this mind blowing stuff. Following are my ideas which i hope to take up for my minor and major projects. Also i plan to write some paper for IEEE Computer Society Student Paper Contest. May be i can use these ideas.
1. open DOORS
2. Drishti
3. Hard OS
4. Dispersed Mobile Computing
5. Overlapped Dimensional View
6. Alpha-Numerology
November 2002: More of these ideas have been hitting me from time to time and i find no time to sit down and write notes on them. I just note the titles down in my diary. I am listing those titles here. I hope to work on some or all of them sometime i.e. if i find time for it all. If you find anything interesting mail me about it with your opinion. I have a general tendency to reply to all emails and in that process i ll be able to generate notes for all these ideas. So here is the list.
Cluster based Neural Networks: Connection Oriented links between various computers
Each computer handling several neurals
Layered approach based message passing
Classification of neural, self-reproduction: Specialization and Generalization
Load Division : Optimizations
Distributed / Parallel Speech Synthesis
Various computers handling different groups of syllables
I have got better notes on this in my copy. I ll type it down someday
Analysis of noise by recording same signal with a noise removal mechanism and with normal recording
Automatic run-time volume adjustment
With Context: To match a particular level
Only amplitudes adjustment
Eye (Retina) - based human recognition: The minority report way
Data-driven pitch estimation and modification of pitch and energy to match estimated value
Big database containing multiple manifestations
Derive pitch change ratio or related information for various transitions
Study significance in speech applications
Minority Report User Interface
Touch Surface or Gesture based drag, manipulate and drop interface
Speech Recognition: Instead of single microphone recording, use a multiple sensor based vocal tract recording instrument for more parameters and data driven approach (Analogous to human hearing with 2 ears)
CSIDC Idea possibility: Global Unification of Money (I have got notes on it in my diary)
Mobile phone anti-stealth
Application residing on the phone
Automatically picks up a call from a special dialer number and followed by a stealth code identifier and and starts sending out messages to rescue it self or makes loud noises in public
Cyclic Signatures: Vague idea ... was just thinking of it some day
Data over Voice Channel: I have been thinking on this for quiet some time
Connect 2 PCs by connecting their AUDIO OUT to the others AUDIO IN in both directions. This is a voice grade channel.
Convert the data into a bit stream and add appropiate headers
Now make a WAVE file with n samples per bit and a fixed amplitude for a 1 and no sound and again n samples for zero
Play this wave file to output and hear it at input of the other PC.
Now apply some DSP at receiver end to get back what was sent.
You need to handle problems like synchronization, volume, etc.
Major Project Idea: Learning Agents (Wrote some thing about it in diary)
User interacts with the agents regularly with the motive of teaching it.
Agent learns and builds a knowledge base.
Check Conflicts
Generates questions intelligently
A responsible user teaches the agent with properly formed sentences and if possible, with PoS indicators
Also i was thinking about a NLP module for Deepti and i wrote a few points down
We can include the learning agent mentioned above into the NLP module and make the module intelligent enough to automatically shift from answering mode to questioning mode.
Short-term memory
Sentence Classifier: Declarative, Questions, Explanative, Exclamation, etc.
Keyword Search: Related to PoS tagging and Noun Identification (May be based on Dictionary Lookup)
Sentence Classification can effect generic replies
Phrase Identifier: Proverbs
Free form Structure of Hindi: Use it
More will be coming...