The Auditory front-end was developed entirely using Matlab version 188.8.131.522 (R2014a). It was tested for backward compatibility down to Matlab version 184.108.40.2063 (R2012b). The source code, test and demo scripts are all available from the public repository at https://github.com/TWOEARS/auditory-front-end.
All files are divided in three folders,
containing respectively the documentation of the framework, the source code, and
various test scripts. Once Matlab opened, the source code (and if needed the
other folders) should be added to the Matlab path. This can be done by executing
startAuditoryFrontEnd in the main folder:
As will be seen in the following subsection, the framework is
request-based: the user places one or more requests, and then informs
the framework that it should perform the processing. Each request
corresponds to a given auditory representation, which is associated with
a short nametag. The command
requestList can be used to get a
summary of all supported auditory representations:
>> requestList Request name Label Processor ------------ ----- ------------------- adaptation Adaptation loop output adaptationProc amsFeatures Amplitude modulation spectrogram modulationProc autocorrelation Autocorrelation computation autocorrelationProc crosscorrelation Crosscorrelation computation crosscorrelationProc filterbank DRNL output drnlProc filterbank Gammatone filterbank output gammatoneProc gabor Gabor features extraction gaborProc ic Inter-aural coherence icProc ild Inter-aural level difference ildProc innerhaircell Inner hair-cell envelope ihcProc itd Inter-aural time difference itdProc moc Medial Olivo-Cochlear feedback mocProc myNewRequest A description of my new request templateProc offsetMap Offset map offsetMapProc offsetStrength Offset strength offsetProc onsetMap Onset map onsetMapProc onsetStrength Onset strength onsetProc pitch Pitch estimation pitchProc precedence Precedence effect precedenceProc ratemap Ratemap extraction ratemapProc spectralFeatures Spectral features spectralFeaturesProc time Time domain signal preProc
A detailed description of the individual processors used to obtain these auditory representations will be given in Available processors.
The implementation of the Auditory front-end is object-oriented, and two objects are needed to extract any representation:
- A data object, in which the input signal, the requested representation, and also the dependent representations that were computed in the process are all stored.
- A manager object which takes care of creating the necessary processors as well as managing the processing.
In the following sections, examples of increasing complexity are given to demonstrate how to create these two objects, and which functionalities they offer.