Digital telephony: some whys and hows of monitoring speech quality
While the importance of telephone technology and speech quality is sometimes overshadowed by emphasis on new types of digital voice services (e.g. the post “Interactive voice: that was then, this is now, what’s next?”), your customers, your workforce, and your business rely on the quality of enterprise phone systems.
Ensuring speech quality on digital telephone services
Fortunately there are ways to make sure your voice services (IVR, calls, CTI) are performing optimally and that telecom operators are providing good service levels (SLAs).
Monitoring, measuring, metrics: IVR, voice calls, CTI
How is the voice quality of telephone systems tested? Let’s take three typical use cases:
- For IVR servers, testing involves listening to voice messages, issuing DTMF sequences (“dialing”), sending synthesized speech messages, and listening to DTMF sequences (“answering”).
- For voice calls, communication is tested on standard telephones or softphones.
- Lastly, computer-telephony integration can be monitored, for instance to simulate call center virtual operators with multiple-mode scenarios (for testing web, mobile, etc.).
These three types of use cases all involve bidirectional testing; in other words, placing a call and answering a call or, to put it differently, acting as both caller and callee. To do so, robots run test scenarios (possibly with SMS in addition to synthesized speech) on real telephone equipment, including softphones, at pre-defined frequencies.
They can test short calls to obtain metrics (availability, dial tone, call pick-up time, busy signal, garbled message, disconnection during the call, call set-up time or PDD), as well as long calls (minimum 3 minutes, for stability and other measurements) on any kind of communications network (analog/PSTN, as well as ISDN, GSM, GPRS, 4G).
Standards for speech quality measurement
Speech quality measurements can be made with or without comparison to a standard reference (ip-label’s Newtest voice robots, for instance, measure speech quality against a reference file).
Data about the performance of voice services from the end user’s point of view is collected when robots run scenarios. This data is processed according to a cognitive model. The ITU recommends certain quality assessment models (PESQ, POLQA…), which predict mean opinion scores (MOS) on scale of 1 to 5 for various types of speech distortions in audio transmission systems.
User-perceived quality of voice services
The end user’s perception of the quality of audio transmissions – how well he or she can hear voicemail instructions, for example, or communicate with the help desk – is the crucial criterion for evaluating voice quality.
If you would like further information about testing and monitoring the perceptual quality of fixed and mobile voice services, the experts at ip-label are willing and able to answer your questions.