SoundExpert (SE) is a research project. New audio quality metrics, used for computing ratings, was worked out inside the project. This metrics is based on both quantitative (measured by instruments) and qualitative (evaluated by humans) estimations of difference signal. Resulting ratings of perceived audio quality are located on Infinite Grade Impairment Scale. Certain of the metric's components need additional research and development. Below are most important tasks/problems.
1. Verification listening tests
This task could be considered as top-priority for SE project - verification of SE quality ratings against results of standard listening tests. As strong point of SE testing methodology is ability to assess sound sources with small/very small impairments, exactly such sources have to be used for the verification. This requires meeting tough conditions while conducting standard listening tests because in order to get meaningful results reference-class audio equipment and controlled listening environment have to be used; participation of experienced (trained) listeners is required. Such arrangement is far beyond possibilities of the indie-research project. Furthermore the verification has to be done outside SE obviously.
Sound sources could be as follows:
- Psychoacoustic coding (2-3 different coders at high bitrates)
- Low-pass filtering
- Adding of shaped and non-shaped noise at different levels
Assessment of the sources according to SE methodology could be done under less strict conditions with participation of ordinary listeners. Conclusion about effectiveness of proposed new testing methodology and corresponding audio metrics can be derived from comparison of both standard test results and SE ones.
Less expensive variant of such verification test would be reproducing of some already finished listening test which was properly organized and included sources with small impairments; sound material used for testing has to be available. In this case simple listening tests with participation of non-prepared listeners would be required only.
Numerous preliminary listening tests inside SE during research process showed that there is high probability of success. In the case audio consumers and manufacturers will get powerful and still easy-to-use instrument for measuring perceived audio quality.
SE team will be glad to get any help in conducting such verification procedure from any institution interested. Please contact us for details.