3D Tensor Builder for Waveform Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing test and measurement systems using machine learning struggle to effectively process and analyze long waveforms and multiple S-parameter vectors due to limitations in tensor image construction, which restricts the use of high-performing neural networks and increases engineering time for accurate results.
Innovation Solution
The method involves creating 3D RGB image tensors with three color channels, allowing for the representation of multiple S-parameter vectors and longer waveforms, using a Tensor Builder to split long waveforms into segments, separate real and imaginary waveforms, and detect short patterns, enabling efficient data organization and processing within the constraints of deep learning networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional tensor image construction methods are used, then the system can process basic waveform data, but it cannot effectively handle long waveforms and multiple S-parameter vectors
Solution Approach 1:
The patent segments long waveforms into multiple shorter waveform segments that can be individually processed and fitted into the tensor image structure. This allows the system to handle waveforms longer than the traditional tensor image dimensions by dividing them into manageable pieces that retain the essential signal characteristics while fitting within the network input constraints.
Solution Approach 2:
The patent introduces a new dimension to the traditional tensor image construction by incorporating multiple S-parameter vectors along the third dimension (depth/channel dimension). Instead of being limited to a single waveform representation, the system now can stack multiple S-parameter measurements as separate channels or depth slices, effectively transforming the data structure from 2D to 3D tensor format that accommodates both temporal and parameter diversity.
2Measurement precision
If high-performing neural networks like Resnet18 are used, then prediction accuracy improves, but engineering time increases due to limitations in tensor image construction
Solution Approach 1:
The patent performs preliminary actions by automatically segmenting long waveforms and organizing multiple S-parameter vectors into the proper tensor image format before feeding them to the neural network. This preprocessing step, which includes waveform segmentation, normalization, and structured arrangement, is performed automatically as part of the input pipeline, eliminating the need for manual data preparation and reducing engineering time while maintaining the high accuracy requirements of advanced networks like Resnet18.
3Quantity of substance
If the tensor image structure is expanded to accommodate more data, then processing capability improves, but the constraints of deep learning networks are violated
Solution Approach 1:
The patent implements dynamic tensor image construction where the segmentation strategy and tensor dimensions are adaptively determined based on the input waveform characteristics and the specific requirements of the deep learning network. The system dynamically adjusts the number of segments, segment length, and tensor dimension sizing to maximize data utilization while ensuring compatibility with network input constraints, rather than using fixed static dimensions.
Data Source
AI summary
A test and measurement instrument includes a port to connect to a device under test (DUT) to receive waveform data, a connection to a machine learning network, and one or more processors configured to: receive one or more inputs about a three-dimensional (3D) tensor image; scale the waveform data to fit within the 3D tensor image; build the 3D tensor image; send the 3D tensor image to the machine learning network; and receive a predictive result from the machine learning network. A method includes receiving waveform data from one or more device under test (DUT), receiving one or more inputs about a three-dimensional (3D) tensor image, scaling the waveform data to fit within the 3D tensor image, building the 3D tensor image, sending the 3D tensor image to a pre-trained machine learning network, and receiving a predictive result from the machine learning network.