Chromatogram Peak Picking via Intensity Normalization and Device-Specific Criteria

The patent applies parameter changes by normalizing chromatogram intensity scales to a standard range (e.g., 0-1) before feeding data to the neural network. This transformation of the intensity parameter eliminates variations caused by different devices or conditions, allowing the automated peak picking system to maintain high accuracy across diverse chromatograms while preserving full automation.

The patent implements preliminary action by collecting and storing device-specific criterion data during a setup phase before actual peak picking operations. The neural network is trained in advance using learning data that incorporates device-specific characteristics (e.g., gas chromatograph vs. liquid chromatograph variations). This pre-training enables the system to adapt to specific device types while maintaining high precision, resolving the contradiction between adaptability and measurement precision.

The patent replaces traditional mechanical or manual peak picking methods with a neural network-based automated system. The neural network processes chromatogram data to identify peak start and end points, automatically determining peak boundaries even in complex scenarios with numerous peaks or invisible noise. This substitution maintains ease of operation through full automation while significantly improving reliability by leveraging the neural network's ability to detect subtle patterns that traditional methods miss.