Audio quality statistics events (AQS)

Verint customers can validate the audio quality of recorded interactions. Audio quality can potentially be used to isolate issues in the recording system Process of capturing an employee-customer interaction for playback, screen and dialer-based interactions, and evaluation purposes., network, or associated communications system. Audio quality statistics events are available for real-time or recorded interactions. These scores show interactions with low audio quality, which can help identify issues with hardware or software resources, the network, or VPN connections. In Automated Verification, you can view the audio quality statistics results of your calls associated with the data sources of your organizations and set target levels of service to use as performance benchmarks.

The events are stored in the database, so can affect database storage usage.

Viewing the audio quality statistics of an interaction In Speech Analytics, an interaction represents a single part of the contact between one employee and the same customer. In Text Analytics, an interaction is the communication session between one or more employees and the same customer with a unifying contextual element.

When viewing an interaction in Player Application that displays recorded interaction data in audio waveform format, which is a color-coded display of the voice energy levels of the employee and customer. The Player also provides metadata about the interaction., you can use the Tags panel to display voice quality statistics of an audio recording. Audio quality is measured as a score on a scale of 0% to 100%. The higher the percentage, the better the quality of the recording. Scoring takes place for both agent and customer sides, and at the segment level.

Audio quality scores calculated before V15.2 2021R1 are listed as Data Events and are stored and displayed per inum.

Audio quality statistics events in the player

Event scoring

Scoring takes place for both agent and customer sides, and at the interaction level. The scores for an interaction are based on the weighted average scores for each segment in the interaction and its duration. By default, segments less than 15 seconds are not included in the score; however, you can change the duration on the Data Source settings page. The Minimum Recording Duration for AQS setting is available in Advanced mode.

Score	Quality
80 – 100%	Good
50 – 79%	Fair
< 50%	Poor

Audio Quality event descriptions

Event	Description
RTP loss	Losing RTP packets can cause missing audio segments, which degrade voice quality.
SRTP decryption errors	Decryption errors cause silence in recording instead of the decoded voice or audio.
Decoding errors	Decoding errors cause silence in recording instead of the decoded voice or audio.
Media mixing errors	Discarded late frames due to stream synchronization. Some errors are normal; however, too many errors can cause dropping the voice of one or all participants.
Volume	Detects whether the average volume is below a specific threshold. When the average volume is below a specific threshold, the voice cannot be heard.
Silence	Too much silence is a recording issue. Silence can be caused by network errors and media processing issues. The amount of silence can be different for certain use cases In Risk Management, use cases to group interactionss according to the needs of the enterprise. Interactions can reside in multiple cases simultaneously. where long silence is typical, such as trader voice open lines.
Noise	Too much noise can severely degrade voice quality. Noises that affect the original call are caused by: Endpoints with a poor acoustic environment. Endpoints with bad quality devices. Media decoding or decryption errors during recording.
Beeps and clicks	Beeps or clicks are considered decoding issues. Too many beeps or clicks can affect the ability to understand the recording.
Amplitude sharp changes	Speech characteristically has words or phonemes that fade softly. A quick or sharp change in speech signals processing issues, such as dropped voice frames or fragments.
Unnatural silence	Unnatural silence is: Silence followed by sharp amplitude changes. A sign of dropped voice frames or fragments caused by media processing and network issues.
Waveform envelope variance	A histogram is built based on amplitude delta ranges, whose variance shows a statistical correlation to speech. Too much deviation from this model is scored as an unnatural voice, which indicates media processing or network issues.
Net speech	Quantity of net speech, in milliseconds
Signal-to-Noise ratio	A comparison of the level of a desired signal to the level of background noise. For a reliable connection, the signal level has to be significantly greater than the noise level.
Crosstalk	Interference from other channels that results in random signals or sounds in the audio.
Overall	The overall score for the interaction, calculated from the values of the individual statistics.