SRA performs spectral and roughness analysis on user-submitted 250- to ~1000ms-long portions of sound files (.wav/.aif formats). Spectral analysis incorporates an improved STFT algorithm [Fitz,K. and Haken,L. (2002). "On the use of time-frequency reassignment in additive sound modeling,"
J. Aud. Eng. Soc. 50(11): 879-893] and automates spectral peak-picking using the Loris open source C++ class library [Fitz and Haken (CERL Sound Group)]. Users can manipulate three spectral analysis/peak-picking parameters: analysis bandwidth, spectral-amplitude normalization, and spectral-amplitude threshold. Instructions describe the parameters in detail and suggest settings appropriate to the submitted files and questions of interest. The spectral values obtained from the analysis enter a roughness estimation model [Vassilakis,P.N. (2005). "Auditory roughness as a means of musical expression,"
Selected Reports in Ethnomusicology 12: 119-144], outputting roughness values for each individual sine-pair in the fileís spectrum and for the entire file. The roughness model quantifies the dependence of roughness on a sine-pairís (a) intensity (combined amplitude of the sines), (b) amplitude fluctuation degree (amplitude difference of the sines), (c) amplitude fluctuation rate (frequency difference of the sines) and (d) register (lower sine frequency). Presentation of the roughness estimation model and the online tool will be followed by a discussion of research studies employing it and an outline of future possible applications. [Supported
by DePaul University and Eastern Washington University. Programmed by K.Fitz] [see also "SRA