An efficient approach for Residence Time Distribution signal processing and identification

This paper presents a proposed approach for Residence Time Distribution (RTD) signal identification. The RTD signals were obtained from measurements carried out using radiotracer technique. In this approach, firstly signal processing is performed for background correction, radioactive decay correction, starting point correction, filtering and signal extrapolation. After signal processing the Mel Frequency Cepstral Coefficients (MFCCs) and polynomial coefficients features are extracted from the signal or from one of its transforms. Discrete wavelet Transform (DWT), Discrete Cosine Transform (DCT) and Discrete Sine Transform (DST) are tested and compared for efficient features extraction. The neural networks are used for matching the extracted features. The proposed approach is tested by original RTD signal and in the presence of noise before and after performing the signal processing. The experimental results show that the proposed approach with features extracted from the DCT of the RTD signal after signal processing is the most robust and reliable in RTD signal identification.