Feature Extraction for Transient Chemical Sensor Signals in Response to Turbulent Plumes: Application to Chemical Source Distance Prediction

This paper describes the design of a low-pass differentiator filter with linear phase and finite impulse response (FIR) for extracting transient features of gas sensor signals (the so-called bouts) which are relevant for accurately estimating the source-receptor distance in a turbulent plume. Our current proposal addresses the shortcomings of previous 'bout' estimation methods, namely: (i) they were based in non-causal digital filters precluding real time operation, (ii) they used non-linear phase filters leading to waveform distortions and (iii) the smoothing action was achieved by two filters in cascade, precluding an easy tuning of filter performance. The presented filter preserves the signal waveform in the bandpass region for maximum reliability concerning both bout detection and amplitude estimation. Thanks to its FIR design, the filter can be implemented with non-recursive structures, thus being inherently stable and allowing an easy algorithmic implementation and optimization. As a case study, we apply the proposed filter to predict the source-receptor distance from recordings obtained with a metal oxide (MOX) gas sensor in a wind tunnel. We demonstrate that proper tuning of the proposed filter can reduce the prediction error to 8 cm (in a distance range of 1.45 m) improving previously reported performances in the same dataset by a factor of 2.5. The performance of bout-based features are also benchmarked against traditional source-receptor distance estimators such as the mean, variance and maximum of the response. We also study how the length of the measurement window affects the performance of different signal features and how to tune the filter parameters to make the predictive models insensitive to wind speed. A MATLAB implementation of the proposed filter and all analysis code used in this study is provided.