System, devices and methods for predicting the perceived spatial quality of sound processing and reproducing equipment
The present invention relates to a method and corresponding system for predicting the perceived spatial quality of sound processing and reproducing equipment. According to the invention a device to be tested, a so-called device under test (DUT), is subjected to one or more test signals and the response of the device under test is provided to one or more means for deriving metrics, i.e. a higher-level representation of the raw data obtained from the device under test. The derived one or more metrics is/are provided to suitable predictor means that “translates” the objective measure provided by the one or more metrics to a predicted perceived spatial quality. To this end said predictor means is calibrated using listening tests carried out on real listeners. By means of the invention there is thus provided an “instrument” that can replace expensive and time consuming listening tests for instance during development of various audio processing or reproduction systems or methods.
1. A method for single-ended (unintrusive) prediction of perceived spatial quality of sound processing and reproducing equipment, devices, systems or methods (abbreviated DUT (Device under test)), the method of prediction comprising the steps of:
providing a DUT, a spatial sound reproduction quality or reproduction of which is to be tested;
providing one of a test signal or a transcoded test signal, where the test signal is transcoded to a format appropriate for the DUT to thereby obtain the transcoded test signal;
providing said test signal or said transcoded test signal to said DUT;
measuring or recording one or more reproduced or processed signals from said DUT;
applying one or more metrics to said one or more reproduced or processed signals, where said one or more metrics is/are designed for providing a physical measure of either said spatial quality as a holistic quantity or for providing physical measures of specific auditory attributes related to said spatial quality;
during a calibration procedure establishing a relationship or correlation between said physical measure(s) and spatial quality assessments or ratings obtained from listening tests carried out on real listeners;
applying said relationship or correlation to the output from one or more of said metrics thereby to obtain a prediction of the perceived spatial quality (holistic or relating to specific spatial attributes) provided by said DUT.
2. A method according to claim 1 , wherein said test signal is a 5 channel de-correlated pink noise signal.
3. A method according to claim 1 , wherein said test signal is pink noise bursts, pair-wise constant power panned from 0° to 360° in 10° increments.
4. A method according to claim 1 , wherein said test signal consists of 8 talkers surrounding a listener at equal angles of 30 degrees.
5. A method according to claim 1 , wherein said test signal contains only anechoic (dry) recordings of speech.
6. A method according to claim 1 , wherein said test signal contains only very reverberant counterparts of recordings of speech.
7. A method according to claim 1 , wherein said test signal is created in a specific channel format corresponding to the format of the system under test.
8. A method according to claim 1 , wherein said transcoding is the transcoding that is required in order to be able to use a test signal comprising a universal directional encoding.
9. A method according to claim 8 , wherein said universal directional encoding is a high order spherical harmonics for driving a standard 5.1 surround sound loudspeaker set-up.
10. A method according to claim 1 , wherein said metrics comprise a “hierarchy” of metrics, where low-level metrics are derived directly from raw data and higher-level metrics derive the final objective measure from a set of low-level metrics.
11. A method according to claim 1 , wherein said relationship or correlation comprises look-up tables, artificial Neural Networks and regression models.
12. A method for double-ended (intrusive) prediction of perceived spatial quality of sound processing and reproducing equipment, devices, systems or methods (abbreviated DUT (Device under test)), the method of prediction comprising the steps of:
providing an equipment, device, system or method (DUT), a spatial sound reproduction quality or reproduction of which is to be tested;
providing one of a test signal or a transcoded test signal, where the test signal is transcoded to a format appropriate for the equipment, device, system or method (DUT) to thereby obtain the transcoded test signal;
providing said test signal or said transcoded test signal to said equipment, device, system or method (DUT);
measuring or recording one or more reproduced or processed signals from said equipment, device, system or method (DUT);
applying one or more metrics to said one or more reproduced or processed signals, where said one or more metrics is/are designed for providing a physical measure of either said spatial quality as a holistic quantity or for providing physical measures of specific auditory attributes related to said spatial quality,
providing either the test or the transcoded test signal to a reference equipment, system, device or method;
measuring or recording one or more reproduced or processed signals from said reference equipment, device, system or method;
applying one or more metrics to said one or more reproduced or processed signals from the reference equipment, device, system or method, where said one or more metrics is/are designed for providing a physical measure of either said spatial quality as a holistic quantity or for providing physical measures of specific auditory attributes related to said spatial quality;
providing output signals from said metrics applied on said DUT and on said reference equipment, system, device or method, respectively;
carrying out a comparison or forming a difference between the outputs from the metrics from said DUT and said reference equipment, system, device or method, respectively, said comparison or difference forming a relative measure for predicting a difference between spatial attributes of the DUT and the reference equipment, system, device or method;
during a calibration procedure establishing a relationship or correlation between said relative measure and spatial quality ratings obtained from listening tests carried out on real listeners;
applying said relationship or correlation to the output of said comparison or difference, thereby to obtain a prediction of the perceived spatial quality difference (holistic or relating to specific spatial attributes) between said DUT and said reference equipment, system, device or method.
13. A method according to claim 12 , wherein said test signal is a 5 channel de-correlated pink noise signal.
14. A method according to claim 12 , wherein said test signal is pink noise bursts, pair-wise constant power panned from 0° to 360° in 10° increments.
15. A method according to claim 12 , wherein said test signal consists of 8 talkers surrounding a listener at equal angles of 30 degrees.
16. A method according to claim 12 , wherein said test signal contains only anechoic (dry) recordings of speech.
17. A method according to claim 12 , wherein said test signal contains only very reverberant counterparts of recordings of speech.
18. A method according to claim 12 , wherein said test signal is created in a specific channel format corresponding to the format of the system under test.
19. A method according to claim 12 , wherein said transcoding is the transcoding that is required in order to be able to use a test signal comprising a universal directional encoding.
20. A method according to claim 19 , wherein said universal directional encoding is a high order spherical harmonics for driving a standard 5.1 surround sound loudspeaker set-up.
21. A method according to claim 12 , wherein said metrics comprise a “hierarchy” of metrics, where low-level metrics are derived directly from raw data and higher-level metrics derive the final objective measure from a set of low-level metrics.
22. A method according to claim 12 , wherein said relationship or correlation comprises look-up tables, artificial Neural Networks and regression models.