Phase characterization of soundscapes
Abstract
Signal processing of soundscapes remains far from the auditive perception. It is particularly true when sounds recorded inside railway stations, are compared with the most accurate spectrographic representations. Such architectural realizations provide a soundscape between closed and opened spaces. As microphones used to discriminate sound sources as well as low level sounds use diffractions properties of radiation, as time discrimination between near sounds is efficient through time domain perception, as filtering often introduces phase distorsions, it has been decided to explore a new kind of signal representation based on the phase of the Short Time Fourier Transform. After a basic presentation of the numerical tool named SAFIR for Spectrography in Amplitude and Frequency, Instantaneous and Reallocated, examples will be given that first show the differences between closed and opened spaces, then the differences between the soundscape inside a train and inside a station. As this work has been realized with sounds recorded for a later musical realization that deals to reconstruct both a new realistic or imaginary coherent soundscape, it has been essential to extract the pertinent, coherent, details as magnification of time or space opposed with the incoherent urban noise. Finally SAFIR will be compared with the composer work.