Teraquant has been Malden’s partner in the USA and Canada since 2005. Bringing 23 years of experience in speech quality assessment, we work closely with our many customers to ensure that Malden’s accuracy and excellence is made available to your wide range of voice and audio test needs.

If you want to distinguish your product with crystal clear audio and inspiring experience for your customers, Teraquant is the best partner to help you measure and improve your product.

DSLA3 Network

Making Accurate Voice & Audio Quality Measurements

Opale Systems produces the most accurate measurement of audio perceptual measurements available worldwide today such as:

  • PESQ (ITU-T Recommendation P.862)
  • PEAQ – ITU-R Recommendation BS.1387
  • POLQA® (ITU-T Recommendation P.863)

We also provide the most accurate measurement of one-way delay for a voice path available, based on the delay of the audio and not just delay of packets across an IP network.

Malden, now Opale, was the first to launch the new POLQA standard on a test platform when it was introduced on the MultiDSLA speech-quality tool in 2011. Opale Systems MultiDSLA also provides unique analytics, drilling deep into the POLQA metrics to provide insight on what is going wrong when you get poor scores.

The DSLA3 is used for detail analysis and characterization of all audio including voice and music. Applications range from silicon chip manufacturers to leading mobile device vendors and wearable technology. It allows for a productive customer and agent experience in the call center in operational networks, ensuring that an “upsell” opportunity is never missed.

Learn more about MultiDSLA.

Perceptual Metric Assessment & MOS

Objective perceptual metrics process signals and provide a score, typically from 1 to 5, indicating how we human beings perceive the quality of the processed sound. A pure reference file is the input to the System Under Test (SUT), e.g., a network or the Device Under Test (DUT). The output is called the degraded file.

Subjective Testing vs. Objective Testing

Subjective testing is human beings giving their opinion. Objective testing is when a computer runs a complex algorithm to analyze the speech, applies some processing and shaping to model the human ear, and enhances certain aspects of the audio which we humans are more sensitive to.

The subjective scale, as shown, is from 1 to 5 and is called Mean Opinion Score (MOS). It includes the mean opinion scores of all people involved in the test. An 8 to 12 seconds speech file is played to the listener after which they are asked to grade the quality of the audio clip. This is the reason why in objective test testing such as PESQ, an 8 second speech file must be used.

MultiDSLA is unique in that it allows you to analyze the speech quality of each frame of voice, e.g., 20 milliseconds, so you can identify exactly what artifacts are causing your codec challenges or what precisely led up to the voice quality problem in the System Under Test.

The audio clip may be pure, good quality audio or it may contain impairments and artifacts. These results are pushed up into a central distributed database. We have around 250,000 audio scores on this database against which we can compare our objective algorithms and ensure they have high correlation to human experience.

MultiDSLA uniquely gives 97% correlation to the database of subjective test results within a 95% confidence interval. This is MultiDSLA’s supreme accuracy.

Voice Quality Algorithm History

PESQ and POLQA are algorithms used for voice quality assessment. PEAQ is used to assess audio or music or media quality. All three are perceptual algorithms which means they are intended to correlate to the human perception of the quality of the audio.

PSQM was first released in 1998. This was superseded in 2003 by PESQ. The main difference between the two was that PESQ allowed for measuring voice quality across an IP network where there might be jitter. This meant that the time alignment between the reference signal and the degraded signal must be readjusted during the 8 seconds of audio speech.

What is POLQA?

Perceptual Objective Listening Quality Analysis (POLQA) is the latest (2011) objective measurement for speech quality standardized (ITU-T P.863). POLQA is an enhancement to the PESQ algorithm (ITU-TP.862) introduced in 2003 which allowed perceptual speech quality measurements to become a mainstream method to evaluate end-to-end speech clarity. Perceptual speech quality measurements are needed when speech is compressed such as in a voice codec (e.g., G.729, G.711, G.722, AMR-WB, EVS, etc.) and then transported over wireless or IP-based networks.

The POLQA algorithm was introduced to deal with some of the weaknesses inherent in PESQ, as well as to address wideband telephony or HD voice (i.e., voice communication above 3.4kHz).

The principle is similar to PESQ — comparing the clean reference file with the degraded file from the network. But the test methodology is different. POLQA has two models, one for narrowband and one for super-wideband. The POLQAv3 (download POLQAv3 multi DSLA Datasheet) goes to full-wideband, the full range of the human ear, and can be used for any network from 300–24,000Hz analog bandwidth. The narrowband POLQA algorithm is for legacy telephony on networks up to 3.4kHz and provides compatibility with PESQ NB and narrowband subjective test data.

MultiDSLA makes it easy to use POLQA. Malden’s long experience in the application of speech quality metrics, inherited by Opale, ensures that you will be getting useful results as soon as the POLQAv3 option is enabled on your MultiDSLA system.

Why Is POLQA Useful?

The main driver for the new POLQA standard was wideband telephony, sometimes called HD-voice as defined by the G.722 and G.722.1 (wideband AMR). In addition, POLQA analyzes speech to the limit of the human ear (24kHz analog bandwidth). The POLQA algorithm supports measurements from 300–3.4kHz including traditional PSTN telephony (when set to the narrow band scale), wideband (WB up to 7kHz), super-wideband (SWB, 50–14kHz), and fullband (FB, 50–24kHz).

Why is Accuracy Important?

  • Accuracy avoids the cost of chasing and repairing a problem which has no impact on quality or user experience.

  • Accuracy allows you to pin-point the real cause of bottlenecks in your system.

  • Accuracy prevents you from failing a good product.

Uses of Perceptual Quality Testing

You should consider testing with perceptual quality algorithms such as PESQ and POLQA if you are testing any device that delivers audio. Perceptual quality testing is the essential “without which nothing else works” test to show when you have changed any parameters in your product or your network service. This ensures that you haven’t broken anything related to the fundamental purpose of the release or upgrade.

Field Test with Outdoor Maps

Test multiple “Fixed Line to Mobile” or “Mobile to Mobile” communications simultaneously. GPS localization and synchronization. Post process mapping data into Google Earth.

Cloud Based Drive Testing
Google Earth

Audio Quality as a Competitive Advantage

In addition, many leading audio technology vendors compete on the quality of the voice over their network or media over their device.

Technologies to be Tested

  • are using either wideband or super-wideband codecs (7–24kHz analog bandwidth), sometimes known as high-definition audio
  • use codecs that employ time-warping (e.g., CDMA or EVS)
  • work with LTE, 5G, or EVS terminals and/or networks
  • have a network that uses AMR codecs
  • are testing through acoustic interfaces
  • are concerned about voice quality enhancement and noise reduction devices/processes

Device Testing

  • voice codec performance
  • compare different voice compression algorithms and implementations
  • determine the bandwidth required to achieve a given level of voice quality
  • evaluate packet loss concealment mechanisms
  • determine capacity of devices such as gateways sharing DSP resources between voice streams

Network Test and Monitoring

  • isolate network elements contributing to poor voice quality
  • baseline current voice quality to set realistic goals for VoIP migration
  • monitor voice SLAs
  • provide objective acceptance criteria
  • eliminate good networks from bad networks

Example Devices to be Tested

  • any device with a voice codec (e.g., VoIP endpoint or transcoder)
  • any device that delivers audio such as:
  • telephone handsets
  • bluetooth speakers
  • softphones
  • mobiles – cellular and Wi-Fi
  • smartphones and smart speakers
  • wearables
  • Wi-Fi Access Points
  • walkie-talkie radios
  • headsets with electronic processing such as noise cancellation
  • cell phones
  • wireless networks
  • VoIP networks
  • MCPTT Servers
  • Transcoders/SBCs
  • IP PBXs
  • VoIP gateways

If you would like to know more about the subjects in this article, please get in touch as below.

Celebrating 20 Years of Excellence

Two decades of passionately working side-by-side with telecom leaders & business enterprises.
See how Teraquant can improve your business communications today!