Hi,
Finally, all the things we discussed before have been put together as a new software product! Sevana AQuA – Audio Quality Analyzer – Automated Voice Quality Testing Software
Test results are available here: AQuA voice quality testing results: percentage of audio similarity, MOS, PESQ
Screenshots of the software one can check here: AQuA voice quality testing screenshots: help, how to use, examples, reasons for voice quality loss
With AQuA you can get three different scores corresponding to your audio quality: percentage of audio files similarity, MOS and PESQ values.
AQuA is a simple but powerful tool to provide perceptual voice quality testing and audio file comparison in terms of audio quality. This is the easiest way to compare two audio files and test voice quality between original and degraded files.
Perceptual voice quality measurement is a way to calculate MOS values and typically one must use quite expensive hardware/software solutions provided by the World leaders in this technology or (in most of the cases) their resellers, unless one has in-house developed P.862/P.863/P.563 implementation.
Another option is measuring the so-called R value. R-value is a number, or score, th…
Hi,
Finally, all the things we discussed before have been put together as a new software product! Sevana AQuA – Audio Quality Analyzer – Automated Voice Quality Testing Software
Test results are available here: AQuA voice quality testing results: percentage of audio similarity, MOS, PESQ
Screenshots of the software one can check here: AQuA voice quality testing screenshots: help, how to use, examples, reasons for voice quality loss
With AQuA you can get three different scores corresponding to your audio quality: percentage of audio files similarity, MOS and PESQ values.
AQuA is a simple but powerful tool to provide perceptual voice quality testing and audio file comparison in terms of audio quality. This is the easiest way to compare two audio files and test voice quality between original and degraded files.
Perceptual voice quality measurement is a way to calculate MOS values and typically one must use quite expensive hardware/software solutions provided by the World leaders in this technology or (in most of the cases) their resellers, unless one has in-house developed P.862/P.863/P.563 implementation.
Another option is measuring the so-called R value. R-value is a number, or score, that is used to quantitatively express the subjective quality of speech in communications systems, especially digital networks that carry voice over IP traffic, or for which VoIP service is under consideration. The R-value score, which is used in conjunction with voice testing processes, can range from 1 (worst) to 100 (best), and is based on the percentage of users who are satisfied with the quality of a test voice signal after it has passed through a network from a source (transmitter) to a destination (receiver).
However, many companies that demand voice quality measurement whether cannot afford available solutions or are not satisfied with the measurements they provide. These are laboratories involved in codec development, telecom and voice/audio systems providers, VoIP providers, companies that develop network and other Quality of Service (QoS) monitoring solutions, IT vendors and consulting companies that provide system design, testing, discovering what causes failures and voice quality losses. For these companies Aqua – software that can be distributed as a DLL library for Windows systems and is available for Linux and in Java is an easy and inexpensive way to introduce voice quality testing and measurement to their systems in a short term and at a low cost.
Our software has these main features that are demanded and very useful for many companies:
– based on research of such scientists as Zwicker, Fletcher, Sapozhnikov, Sorokin we can create a test audio file containing the so-called critical bands that are important for human ear perception
– we can quickly compare two voice files to obtain estimation on how different they are quality wise (based on the research of the same scientists)
– after the files are compared we provide reasons for quality loss with quantative estimations like:
Duration distortion.
Audio shrinking corresponds to 6.33 percent.
Audio signal activity mistiming (unsynchronization) is 9.90 percent.
Corrupted signal spectrum.
Overall specral energy distortion approaches 26.89 %
Vibration along the whole spectrum [-12.14, 14.75] %
Significant distortion in low frequencies band.
Energy distortion approaches 10.66 %
Spectrum vibration in low frequency band [-4.48, 6.18] %
Significant distortion in medium frequencies band.
Energy distortion approaches 12.52 %
These features allow obtaining some kind of “R value” automatically by comparing voice files using our methods. We can also provide on the fly channel testing if one streams our speech model audio and then compares it with what is received at the destination and we are also able to scale down our values (that are between 0 and 100% of signal similarity quality wise) to PESQ or/and MOS values, what is implemented in the latest version. Speech model is also used to test voice codecs quality f.e. to choose the best for one’s voice system. And on top of that we can actually provide a quantative feedback with numerical data describing what were the reasons for quality loss.
Our approach is new and not related to ITU standards, what maybe disadvantage to some companies, because although even if P.862 fails it still corresponds to a World telecom standard; we know about that and are working on becoming a standard (we welcome partners and investors for these activities), but our approach is also DIFFERENT to the existing ones and provides very good results. For example: http://www.microtronix.ca/pesq-disc.html
Our software does catch that significant quality loss, which occurs when the file is equalized such that there is far less low frequency and high frequency energy when compared to the original file.
Our software consumes about 2 times less computational power than PESQ implementations (can serve 10 ports instead of 5 on the same hardware), can easily be integrated, and our pricing policy is much more attractive than of those who offer ITU P.863 or similar standards.