Object Recognition via Capacitance-Based Spike Signal Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing spike neuron networks (SNNs) face challenges in recognizing objects with high resolution images and static objects, as they require larger input processing devices and struggle with processing static images.
Innovation Solution
The proposed object recognizing device includes an electrode array with drive and sense electrodes, a sensing circuit that generates a pulse train representing spatial and material information of an object based on mutual capacitances, and a neuron network processor that recognizes the object from the pulse train.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image data is processed with high resolution to improve object recognition accuracy, then recognition precision is improved, but device size and complexity increase
Solution Approach 1:
The patent extracts only the essential features from images - specifically spatial information (position of objects) and material information (dielectric constant). Instead of processing entire high-resolution images, the system extracts these key features and encodes them into compact spike signals, thereby maintaining recognition accuracy while dramatically reducing device complexity and size
Solution Approach 2:
The patent inverts the traditional approach by not processing images directly through complex neural networks. Instead, it converts image features into spike signals that naturally encode both spatial and material information, allowing the neuron network processor to recognize objects with simpler architecture and reduced computational requirements
2Speed
If dynamic vision sensor is used to process images, then processing speed is improved, but static objects cannot be recognized
Solution Approach 1:
The patent creates a universal input system that can handle both static and dynamic objects through the same electrode array and sensing circuit. The mutual capacitance measurement approach works for any object regardless of motion state, making the system multi-functional while maintaining fast processing speeds through direct electrical measurement rather than sequential image capture
3Productivity
If mutual capacitance values are directly processed, then processing efficiency is improved, but spatial and material information separation is difficult
Solution Approach 1:
The patent segments the mutual capacitance information into two distinct components: spatial information (encoded in the timing and pattern of spike signals based on electrode geometry and object position) and material information (encoded in the amplitude or frequency characteristics based on dielectric constant). This segmentation allows efficient processing while preserving and separately utilizing both types of information in the spike signal representation
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This solution enables efficient object recognition with low-resolution data, effectively handling both static and dynamic objects by converting mutual capacitance values into spike signals for processing by a spike neuron network.
Implementation Method 1
generating a pulse train representing spatial information and material information on an object positioned over the electrode array based on mutual capacitances formed between the plurality of drive electrodes and the plurality of sense electrodes
Data Source
AI summary
A method and a device for recognizing an object are disclosed. The object recognizing device includes an electrode array including a plurality of drive electrodes and a plurality of sense electrodes; a sensing circuit configured to generate a pulse train representing spatial information and material information on an object positioned over the electrode array based on mutual capacitances formed between the plurality of drive electrodes and the plurality of sense electrodes; and a neuron network processor configured to recognize the object positioned over the electrode array based on the pulse train.