The purpose of the Two!Ears Auditory Machine Learning Training and Testing Pipeline (AMLTTP) is to build and evaluate models for auditory sound object annotation and assigning attributes to them. The models are obtained by inductive learning from labeled training data. The framework is tightly coupled with the Two!Ears system.
While the pipeline is designed with flexibility in mind and is extendable to new target attributes, data features, or model and training algorithms, it so far serves the specific purpose of training and evaluation of block-based auditory object-type, object-location, and number-of-sources classifiers using data from simulated auditory scenes generated within the same framework.
In [1], we have developed and analyzed robust binaural sound event detection models using AMLTTP.
[1] Trowitzsch, Ivo, et al (2017). Robust detection of environmental sounds in binaural auditory scenes, IEEE/ACM Transactions on Audio, Speech, and Language Processing 25(6), 1344-1356.
The AMLTTP makes use of other software modules of the Two!Ears Computational Framework. You will need to download
- https://github.com/TWOEARS/blackboard-system
- https://github.com/TWOEARS/auditory-front-end
- https://github.com/TWOEARS/binaural-simulator
- https://github.com/TWOEARS/SOFA
- https://github.com/TWOEARS/main
- https://github.com/TWOEARS/stream-segregation-training-pipeline.git (if you want to use spatial segregation in AMLTTP)
In your "main"-directory, please first edit TwoEarsPath.xml to point to your respective directories.
Once Matlab opened, the source code folders need to be added to the Matlab path. This will be accomplished by executing the following commands in:
>> addpath( '<path-to-your-TwoEars-Main-directory>' )
>> addpath( '<path-to-your-AMLTTP-directory>' )
>> startAMLTTPThe complete functionality of the AMLTTP will be explained in detail in the accompanying Online user manual.
The AMLTTP is developed by Ivo Trowitzsch and Youssef Kashef from Technische Universität Berlin.
If you use this software, please cite as
Ivo Trowitzsch, Youssef Kashef, and Klaus Obermayer. (2019). Auditory Machine Learning Training and Testing Pipeline: AMLTTP v3.0 [Software]. Zenodo. http://doi.org/10.5281/zenodo.2575086
The AFE is released under the BSD 2-Clause license.
This project has received funding from the European Union’s Seventh Framework Programme for research, technological development and demonstration under grant agreement no 618075.
