Signal Classification Neural Network Training via Virtual Waveform Primitives
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing volume of data from signal analyzing apparatuses, such as digital oscilloscopes, makes signal analysis time-consuming and prone to errors, making it impractical to effectively process and display acquired signals.
Innovation Solution
A method is introduced to generate virtual waveform primitives representing specific signal levels and edges, forming a training data set for a signal classification neural network, which is then used to classify signals, allowing for efficient and accurate analysis by reducing the complexity of signal processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Duration of action of moving object
If the memory size of the acquisition memory is increased to record longer signals, then the recording time is increased, but the signal analysis becomes more time-consuming and error-prone
Solution Approach 1:
The patent segments the complex signal analysis task into classification of individual waveform primitives (rising edges, falling edges, high levels, low levels). Instead of analyzing the entire long signal at once, the neural network classifies discrete primitive elements that compose the signal, making the analysis of extended recording times feasible and efficient
Solution Approach 2:
The patent uses virtual waveform primitives as synthetic copies or representations of actual signal features. These virtual primitives are generated through simulation and used to train the neural network, enabling the system to learn from idealized versions of signal patterns without requiring extensive manual analysis of raw data
2Duration of action of moving object
If the memory size of the acquisition memory is increased to record longer signals, then the recording time is increased, but the analysis becomes more error-prone
Solution Approach 1:
The patent replaces manual or conventional algorithmic signal analysis with a neural network-based classification system. This substitution of the analysis mechanism provides more consistent and reliable results, especially for long signals where manual analysis errors would accumulate. The neural network systematically identifies waveform primitives with high accuracy regardless of signal length
Solution Approach 2:
By training the neural network on virtual waveform primitives that accurately represent real signal features, the system learns robust classification patterns. This copying of essential signal characteristics into simplified virtual forms enables the network to reliably identify actual signal features even in complex, long-duration recordings
3Quantity of substance
If conventional signal analysis methods are used on large volumes of acquired data, then all signal data can be processed, but the analysis becomes unpractical and almost impossible
Solution Approach 1:
The patent divides the continuous signal into discrete waveform primitives (rising edges, falling edges, high levels, low levels). This segmentation transforms the overwhelming task of analyzing large volumes of continuous data into the manageable task of classifying individual primitive elements, making the analysis of large datasets practical and systematic
Solution Approach 2:
The patent changes the parameter representation of signals from continuous amplitude-time data to discrete classified primitives. By transforming the data representation into categorized elements with specific parameters (primitive type, position, duration), the system makes large volumes of signal data tractable for automated analysis
Data Source
AI summary
A method for providing a training data set used for training a signal classification neural network is provided. The method includes generating at least one first virtual waveform primitive comprising a predetermined signal level and at least one second virtual waveform primitive comprising a signal edge. The training data set is formed and comprises a predetermined number of generated virtual waveform primitives including first virtual waveform primitives and second virtual waveform primitives. Each virtual waveform primitive comprises a sequence of time and amplitude discrete values. The training data set is used for training the signal classification neural network.