Multi-Sensor Signal/Noise Detection With Synthetic Training Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing neural network-based signal/noise separation methods require large amounts of training data, which are difficult to acquire.
Innovation Solution
A signal/noise determination apparatus and method using a plurality of sensors, a determination model recording section, and a signal/noise determining section, where the determination model is generated through machine learning with hypothetical signal and noise information as training data, allowing easy acquisition of training data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If neural network-based signal/noise separation is used, then determination accuracy is improved, but training data acquisition difficulty increases
Solution Approach 1:
The patent applies preliminary action by pre-generating hypothetical measurement results, hypothetical signal components, and hypothetical noise components through simulation before actual deployment. This creates a predetermined training dataset that eliminates the need for difficult real-world data collection, while still enabling the neural network to learn effective signal/noise separation patterns.
Solution Approach 2:
The patent uses copying by creating hypothetical copies of measurement data through simulation. Instead of requiring actual measured signal/noise pairs which are difficult to obtain, the system generates synthetic training data that replicates the statistical properties and characteristics of real data, allowing the neural network to be trained effectively without access to rare or difficult-to-acquire real-world examples.
2Reliability
If large amounts of training data are required, then machine learning model performance is improved, but data collection time and cost increase
Solution Approach 1:
The system performs preliminary generation of comprehensive training datasets through simulation, creating large volumes of hypothetical measurement results with known signal and noise components. This pre-computed data repository enables reliable model training without the time-consuming process of collecting equivalent amounts of real-world data, as the synthetic data can be generated on-demand through computational simulation.
3Adaptability or versatility
If hypothetical training data is used, then training data availability is improved, but data authenticity may be compromised
Solution Approach 1:
The patent applies parameter changes by systematically varying parameters in the hypothetical data generation process, including different signal amplitudes, noise levels, spatial configurations, and temporal characteristics. This ensures that the synthetic training data covers a wide range of realistic scenarios and maintains statistical properties similar to actual measurements, thereby preserving data authenticity while ensuring broad availability.
Data Source
AI summary
A signal/noise determination apparatus includes a plurality of sensors, a determination model recording section, and a signal/noise determining section. The plurality of sensors measure a signal and a noise. The determination model recording section records a determination model used to determine whether components of results of measurement by the sensors expected with hypothetical signal information and hypothetical noise information are from a signal source or a noise source. The determination model is generated by machine learning with the measurement results, the hypothetical signal information, and the hypothetical noise information as training data. The signal/noise determining section determines whether components of the measurement results are from the signal source or the noise source based on the measurement results and the determination model. The signal information includes the position of the signal source and the signal, and the noise information includes the position of the noise source and the noise.