Technology Opportunities

Neural network architecture for faster, more accurate learning

Highlights

• Software system for faster speech recognition and image processing
• Based on neural network architecture that supports time-warp invariant computation
• More accurate understanding of accented speech or speech in noisy environments
• For integration into telecommunications hardware (e.g. car phones), computing systems (speech to text)

Our Innovation

Neural algorithm for time-warp invariant signal pattern processing and recognition.

Key Features

• Simpler architecture results in faster learning and processing, faster speech recognition
• Neural algorithm is closer to process used by brain, resulting in faster, more accurate processing
• Can also be applied to very rapid image processing

Development Milestones

Ongoing work to test model in a range of more realistic situations in different environments with background noise

The Opportunity

• For use in mobile phone applications – voice dialling, call routing
• Automated call centers
• Computers – appliance control, data mining
• Security and safety applications – aircraft cockpits

Researcher Information

Professor Sompolinsky: neurophysics.huji.ac.il/~haim/
Dr. Robert Gutig: neurophysics.huji.ac.il/~guetig/