|
Sound acoustics |
|
Ingenieurbüro für Nachrichtentechnik und Akustik |
|
Listening test simulation with HASQUE |
|
The simulation of listening tests makes quality evaluation with strongly reduces effort and without from day dependent health feeling of test persons possible.
The HASQUE principle can be adapted to various listening test conditions with the aid of programmable listening test parameters as playback loudness, bandwidth and threshold of acceptance. HASQUE operates without band limitations as e.g. IRS filters for the simulation of telephone quality and evaluates distortions with the aid of psycho acoustic functions even so during speech pauses. Continuous noise sources as e.g. car noise at constant speed or variable noise sources as e.g. babble noise in a restaurant lead to strong perceptible quality differences. HASQUE operates for this reason with an adaptive masked threshold taking time variant and spectral properties of noise sources into account.
|
|
The experience with subjective listening tests of medium and large companies led to valuable realizations for new algorithms, making today a high correlation with subjective listening tests possible.
The correlation between quality measures of subjective listening tests and objective test principles amounts today up to 0.981 for HASQUE (blue) and 0.764 for conventional other tests (grey) respectively.
Figure left: Quality measures of subjective listening tests (green) and objective test principles (blue, grey) of various applications at different bandwidth.
|
|
The results above are based on a collection of test scenarios with following properties: Bandwidth 4 - 8kHz (8 -1 6kHz sample rate) Signal to noise ratio (SNR) 0 - 40dB Noise sources Street Car Kitchen Restaurant Office
The dotted green D - curves indicate the maximum deviation of test persons with different threshold of acceptance to the MOS - curve and can be interpreted as offset to the MOS curve. D‘ indicates the offset free curve with a correlation of 0.991 with the MOS curve at all. Hence the subjective perceived quality differentiations of different test persons are close together and the absolute quality evaluation depends on the threshold of acceptance of the different test persons , but also from the day dependent health feeling of one test person.
This threshold of acceptance can be determined by the listening test parameters of the HASQUE principle making the simulation of different listening tests possible.
HASQUE is implemented in our measurement systems. |