IVA patents real-time video-communication quality assessment

IVA Technologies has received Russian patent No. 2861389 for a “System for assessing video quality degradation”. The patent was registered on 5 May 2026, and the company announced it on 22 June. The technology analyses a video stream after it has passed through the network, detects several classes of distortion and produces an overall transmission-quality assessment.

Unlike a basic availability check, this approach evaluates what a participant actually receives. The network and application can remain formally available while the picture drops frames, freezes or develops decoding artefacts.

What degradation the system assesses

The official announcement identifies:

• reduced image clarity;
• lost frames;
• frozen video;
• decoding artefacts;
• a combined score based on objective metrics.

Analysis takes place after network transmission. That distinction matters because a source may generate a valid stream while degradation appears later through packet loss, jitter, limited bandwidth, an overloaded intermediary or receiver-side decoding errors.

Why availability alone is insufficient

A video service may return 200 OK, maintain an active session and keep its media component online while users see a frozen or fragmented image. Video-conferencing observability therefore needs several layers:

1. signalling and identity-service availability;
2. successful media-session establishment;
3. latency, jitter and packet loss;
4. lost or repeatedly decoded frames;
5. audio and video synchronization;
6. an end-to-end quality score at the receiver.

Our IT infrastructure observability guide explains how to connect metrics, logs and user-visible symptoms.

Where this assessment is useful

The source mentions video conferencing, streaming and online broadcasts. For businesses, it is most important where image distortion can affect a decision rather than just comfort: a remote interview, medical consultation, examination, negotiation or legally significant hearing.

Operational value appears when a quality score is connected to session context such as site, network provider, device type, codec, resolution and route. Teams can then distinguish one participant’s local problem from broad service degradation.

Building the signal into a video SLO

One average score can hide short but critical interruptions. A useful SLO should combine several measures:

• sessions established within the target time;
• minutes without a freeze longer than the defined threshold;
• P95 latency and jitter;
• sessions meeting the acceptable video-quality score;
• time to detect and localise degradation;
• breakdowns by site, network and client version.

Thresholds should be tied to the user journey rather than selected arbitrarily. Our guide to SLIs, SLOs and error budgets describes this method.

What to validate before adoption

• where the stream is measured: sender, media service or receiver;
• which codecs, resolutions and devices are supported;
• how the score responds separately to frame loss, jitter and bitrate reduction;
• whether scores correlate with controlled user tests;
• what telemetry is retained and whether it contains personal data;
• whether a quality event can be tied to a network segment without recording conversation content;
• how planned quality changes are kept from triggering false alerts;
• which responses are automatic and which require an operator.

When a site has multiple links, a quality score can contribute to failover logic but should not become the only trigger. Resilient branch connectivity explains path independence and controlled failover.

What the source does not yet disclose

The public announcement does not provide the scoring formula, independent comparison results, computing requirements, accuracy by distortion class or a date for inclusion in a production product. IVA says the technology may later become available to third-party services, but gives no timetable or commercial terms.

A patent confirms a technical method, not the readiness of a particular deployment. Customer value will depend on measurement quality across real networks, understandable thresholds, telemetry protection and whether the signal helps engineers locate degradation quickly.