An application of wavelet transform for decomposition of millimeter-wave spectroscopic signals
- Purdue Univ. Calumet, Hammond, IN (United States). Dept. of Engineering
- Argonne National Lab., IL (United States)
Millimeter-wave technique, based on rotational energy transitions of molecules, holds promise for remote monitoring of environmentally hazardous effluents from processes. Argonne National Laboratory is developing a millimeter-wave sensor based on active swept-frequency radar technique in the frequency range of 220-320 GHz. Because the line widths of millimeter-wave spectra of molecules at atmospheric pressure are broad ({approximately} 4 GHz half-width at half height), the composite spectrum of multicomponent mixtures of chemicals is generally complex and overlapping. This paper presents an application of discrete wavelet transform for efficient representation and decomposition of millimeter-wave spectral data. A two-layer back propagation neural network is trained using multifrequency wavelet coefficients of the signals as input features and the known composition of different chemicals in the mixture as target output vectors. After training, composition of an unknown mixture of the base chemicals is determined using the wavelet representation of its absorption spectra. Simulated and experimental spectral data were used to test the wavelet transform technique. Accurate values of individual chemical compositions resulted for noise-free laboratory data. In addition, the technique showed more robustness than conventional multivariate techniques under noisy conditions.
- Research Organization:
- Argonne National Lab., IL (United States)
- Sponsoring Organization:
- USDOE, Washington, DC (United States)
- DOE Contract Number:
- W-31109-ENG-38
- OSTI ID:
- 10181969
- Report Number(s):
- ANL/ET/CP--83924; CONF-9405229--1; ON: DE94018366
- Country of Publication:
- United States
- Language:
- English
Similar Records
Millimeter-wave radar sensing of airborne chemicals.
Advanced millimeter wave chemical sensor.