Improvement of Signal to Noise Ratio and Detection Limit for Gas-Photoaoustic-Signal Detection Based on Feature Extraction in Blind Source Separation

Abstract A mixing gas photoacoustic signal is regarded as a weighted sum of feature absorptions by the spacing between adjacent feature wavelengths. A promising technique, employing feature extraction based on blind source separation, is used to demonstrate the extraction of features with hidden within the sampled data and the inverse of content for the detected gas. This photoacoustic signal measurement based on second harmonic has been established to detect multi-spectrum signal in samples from breath ammonia detection in high concentration of CO2 and H2O at atmospheric pressure. The simulation results of data from a real-time experimental with probed by a second harmonic show an improved detection limit of NH3 in simulated breath samples from the 16 ppb to 10 ppb under a higher analysis accuracy and signal to noise ratio. The procedure can process any detectable gas sample with a “transient shape” feature absorption.