Horiba has filed a patent for an analysis device that can analyze a measurement sample based on spectral data. The device includes a correlation data storage portion that stores correlation data between spectral data for a reference sample and its total analysis value. A calculation main unit applies this correlation data to the spectral data obtained from the measurement sample to calculate the total analysis values of the components in the sample. The reference sample consists of a first sample containing multiple components and a second sample containing one or more components from the first sample. The correlation data includes a machine learning model using training data from both the first and second reference samples. GlobalData’s report on Horiba gives a 360-degree view of the company including its patenting strategy. Buy the report here.
According to GlobalData’s company profile on Horiba, optoelectronic biosensors was a key innovation area identified from patents. Horiba's grant share as of September 2023 was 32%. Grant share is based on the ratio of number of grants to total number of patents.
Analysis device for calculating total analysis values of components in a measurement sample
A recently filed patent (Publication Number: US20230296501A1) describes an analysis device that analyzes a measurement sample based on spectral data obtained from that sample. The device includes a correlation data storage portion that stores correlation data showing the correlation between spectral data for a reference sample and the total analysis value of that sample. A calculation main unit applies the correlation data to the spectral data obtained from the measurement sample and calculates the total analysis values of predetermined components in the sample. The reference sample consists of a first reference sample containing the predetermined components and a second reference sample consisting of either one or multiple components from the first reference sample.
The correlation data stored in the device represents a machine learning model, with training data including spectral data and total analysis values for both the first and second reference samples. The second reference sample can either be one or multiple components that make up the predetermined components, components that do not contribute to the total analysis value, or components with a pseudo-correlation to the total analysis value. In one embodiment, the second reference sample is a fuel that generates exhaust gas.
The analysis device can be an FTIR-type device, and the total analysis values of the first and second reference samples can be obtained through measurements performed by an FID analyzer. The device can store correlation data calculated for different types of fuel, allowing the calculation main unit to switch the correlation data applied to the spectral data based on the type of fuel used to generate the measurement sample.
The patent also describes a method of analyzing a measurement sample based on spectral data, where the correlation data is stored and applied to the spectral data to calculate the total analysis values of the predetermined components in the sample. Additionally, a program for the analysis device is provided, which enables the correlation data storage portion and calculation main unit functions.
Furthermore, the patent includes a learning device for analysis that receives spectral data from a reference sample and stores reference sample data containing total analysis values for different reference samples. A correlation calculating portion employs machine learning to calculate a common correlation between the spectral data and total analysis values of each reference sample. A learning method and program for the learning device are also described.
Overall, this patent presents an analysis device and method that utilize correlation data and machine learning to calculate total analysis values of components in a measurement sample based on spectral data. The device can be used in various applications, including analyzing exhaust gas components in different types of fuel.