Bring the Power of POLQA PESQ and PEAQ to your Quality Optimization Process

Nick Fox General

What’s it all about?

The latest additions and improvements to the Opale Systems MetricServer platform bring the power of the ITU standardized metrics POLQA, PESQ and PEAQ to any organization wanting to take full control of its voice and audio quality optimization procedures.  Read on to learn about automated processing using MetricServer with cross-platform Application Program Interface (API) support.

Background: ‘testing’ vs ‘processing’

When we talk about objective techniques such as POLQA, PESQ and PEAQ for evaluating voice and audio quality we often think in terms of testing – that is, using a test system of some type.  For example, the MultiDSLA system fulfils the requirements of a turnkey test system, i.e:

  1. Generating the ‘reference’ speech or music samples
  2. Passing these through the device or system under test
  3. Capturing the ‘degraded’ speech or music
  4. Processing the results

Requirements of Codec developers and others

The configuration we see above is a generic solution, which adapts well to many different environments and technologies, streamlines the process for the user and provides a great deal of control.  It is also an active, real-time solution, generating results each time a test is made.  But particularly in an all-software environment, such as in the development and characterization of a codec for example, a user’s existing setup will often take care of stages 1-3 above.  What this user needs is a way of post-processing existing pairs of reference and degraded recordings, typically in PCM or ‘wave’ file format, to deliver the scores and other data provided by the chosen metric(s).

N.B. In some cases, active testing with the MultiDSLA system may reveal information not generally available using a post-processing method.  This is particularly true for speech/audio delay or latency, which MultiDSLA measures accurately with synchronized nodes.

High processing volume

The volume of processing can be high, since thorough testing of a codec may involve multiple speech or music samples, multiple real or simulated network conditions, multiple combinations of codec parameters, and so on.  And each time a developer makes a change to the codec, the whole process has to be repeated, to determine the effect on voice or music quality.  MetricServer enables this regression testing to be done reliably, consistently and with minimal learning curve and time overheads.

MetricServer solution

Opale’s MetricServer solution was developed to meet the needs of developers and QA professionals working in fields including speech coding, speech processing, Bluetooth streaming, etc., who need to obtain scores from any of the popular metrics mentioned above.  Such requirements usually include the need for automation of the process and this was recognized in the MetricServer concept.

Simple operating model

MetricServer has a simple principle, based on the Input, Process, Output model.  The Input is a pair of reference and degraded recordings, the Process is one or more of the metrics including POLQA, PESQ and PEAQ; the Output represents the results.

This illustration shows the concept, and also shows that Inputs, Processes and Outputs are each represented by one or more profiles, or collections of settings.  These profiles are unlimited in number and may be combined very easily to cover all the file processing requirements of an organization.

The best of all worlds

MetricServer supports use from a console and use via a Python test script.  The system consists of a central server engine, a web browser application and a full-featured API, primarily supporting the Python scripting language.  The web browser application and the API are both clients to the server, and the standard version supports concurrent use of up to five client sessions, with any mix of browser app and API.  The system can be expanded in blocks of five client sessions to support a whole team.  The web browser application facilitates small-scale or large-scale processing of file pairs in its own right, and also allows MetricServer configurations to be created and validated before an automation script is brought into use.

Expanded Python API support

Recent additions to MetricServer include an enhanced API featuring a Python package compatible with 32- and 64-bit Windows, MacOS and Linux, offering great versatility and a high probability of compatibility with existing platforms.

Simple setup and simple scripting

MetricServer’s design and straightforward functional model leaves users free to concentrate on their real tasks. Initial configuration is simple, and the powerful API simplifies the task of scripting.

Like to know more?

Contact Opale Systems to discuss your requirements, learn how you can take back control of voice and audio quality optimization and arrange a no-obligation trial of MetricServer in your own lab.