BIAS standards for signal processing
The conclusive version of the BIAS standards for signal processing is now available!
The signal processing within the BIAS project is restricted to topics related to the analysis of the retrieved digitized sound data. The main parts are to control the quality of data, to extract relevant estimates out of the raw data set, to present the properties of data in an understandable way and to provide results in an appropriate format for sound scape modelling. Pre-processing is a basic step in signal processing to prepare data for data analysis and to assess the quality of the data.
For the quality control of data BIAS use defined pre-processing procedures such as: testing of self-noise, organisation of recorded data, testing of data coverage, testing of file size and file length of the recorded files, testing of non-numerical values in the recorded data files and testing of clipping of the data.
Signal processing of data constitutes the main analysis of the raw data with the aim of extracting the fundamental measures needed for modelling sound propagation and to sum up the statistics. Processing of the sound data includes both filtering of data and data analysis. In the first step, an FFT-analysis is done over consecutive 1-second periods, giving amplitude spectra with a 1 Hz resolution. In the second step, the sound pressure levels (SPL) are calculated in the required 1/3-octave bands over 1 second. The 1-s averages are then further processed to averages of 20 seconds.
Similarly, estimates such as hourly, daily and monthly averages are also calculated based on the 1-s averages. Finally, the annual mean is derived using the 1-s averages. In BIAS monthly and annual arithmetic means and the percentile levels are established for 63, 125 and 2000 Hz as well as for the frequency band from 10 to 10,000 Hz.