Sensor Signal Processing Using Multi-Model Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing signal processing methods for sensors, such as gas sensors, require a long response time to converge, making it difficult to accurately evaluate properties of objects during this period, and previous estimation methods rely on a single model which can lead to high estimation errors if the model is not accurate.
Innovation Solution
A signal processing apparatus that includes an input interface, a prediction circuit to generate predicted values based on converged values, and an estimation circuit to calculate an estimated value using these predictions, allowing for accurate evaluation of sensor output signals before the response time is reached, by utilizing multiple models and error calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single model is used for estimation, then the device complexity is reduced, but the measurement precision deteriorates due to high estimation errors when the model is not accurate
Solution Approach 1:
The patent divides the estimation system into multiple independent models, each trained on different datasets or with different characteristics. Instead of using a single model, the system segments the estimation task across multiple models (first model, second model, etc.), allowing each to contribute to the final estimation result and reducing the risk of high errors from any single model
Solution Approach 2:
The patent combines the output results from multiple independent models through a synthesis process. The estimation results from the first model, second model, and potentially other models are merged to produce a final estimation value, leveraging the strengths of each model while compensating for individual weaknesses
2Measurement precision
If multiple categories and models are prepared to reduce estimation error, then the measurement precision is improved, but the device complexity increases
Solution Approach 1:
The patent creates models with universal applicability that can handle multiple categories and conditions. Instead of preparing separate specialized models for each category, the system uses models designed to be broadly applicable across different scenarios, reducing the number of models needed while maintaining high estimation accuracy across diverse conditions
3Measurement precision
If the converged value is used for evaluation, then the measurement precision is improved, but the loss of time increases due to the long response time period
Solution Approach 1:
The patent performs preliminary estimation actions using multiple models before the sensor output signal has fully converged. By generating estimation results from multiple models during the transition response period and synthesizing these results, the system obtains accurate evaluations ahead of time, without needing to wait for the full response time period to elapse
4Ease of operation
If a single model is used for estimation, then the ease of operation is improved, but the reliability deteriorates when the model accuracy is insufficient
Solution Approach 1:
The patent implements a feedback mechanism where the estimation results from multiple models are synthesized and evaluated. The system monitors the performance and consistency of each model's output, using this feedback to weight or adjust the contribution of each model to the final result, thereby improving overall reliability while maintaining operational simplicity
Data Source
AI summary
Signal processing apparatus includes: an input interface configured to receive an output signal Va(T) from a sensor; a prediction circuit configured to generate, on the basis of a relationship different depending on each of a plurality of converged values Vc, a plurality of predicted values Vb_T2 corresponding to a value of the output signal that would be obtained at a time T2 after a time T1, in a transition response period before a response time period Tr elapses where Tr denotes a response time period required for a value of the output signal Va(T) to become a converged value Vc corresponding to a value P of a parameter representing a certain property of an object to be measured, in accordance with a value Va_T1 of the output signal obtained at the time T1; and an estimation circuit configured to generate, on the basis of the value Va_T2 of the output signal obtained at the time T2 and the plurality of predicted values Vb_T2, an estimated value Pe of a parameter representing the certain property of the object to be measured.