Method and system for compression-based spectral identification
The system compresses unknown spectra into latent space for efficient spectral identification, addressing inefficiencies in large spectral library searches by reducing data processing and storage needs, enabling faster and more accurate chemical composition determination.
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- THERMO ELECTRONICS SCI INSTR LLC
- Filing Date
- 2025-12-19
- Publication Date
- 2026-07-02
AI Technical Summary
Spectral library searches are inefficient due to large library sizes, leading to long search times and high memory requirements, making them unsuitable for local storage on computing devices or scientific instruments.
A system that compresses unknown spectra into latent space using an encoder and matches them with compressed known spectra, reducing data processing and storage needs, while maintaining accuracy through preprocessing and using a decoder for full spectrum retrieval.
Facilitates faster and more scalable spectral identification by reducing computational and storage requirements, improving search speed and accuracy without the need for a local spectral library.
Smart Images

Figure 2026110573000001_ABST