HiCellTek HiCellTek
← Product

MOS SDK

Measure perceived voice and video quality directly on device. Two integrated MOS scoring engines: ViSQOL (Google Research) for voice, ITU-T P.1204.3 for video. ARM64 native, pre-trained models, JNI ready.

ViSQOL (Google Research) ITU-T P.1204.3 MOS 1 to 5 ARM64 Native JNI Ready Pre-trained Models

HiCellTek MOS SDK provides two perceptual quality scoring engines. The voice module uses ViSQOL (Google Research), a spectro-temporal ML model that produces a MOS score from 1 to 5 for VoLTE and VoNR calls. The video module implements ITU-T P.1204.3, a parametric model that evaluates video quality from stream metadata (bitrate, resolution, codec). Both modules are compiled ARM64 native with JNI interface for Android integration. Voice: EUR 4,490/year. Video: EUR 4,490/year. Bundle: EUR 6,990/year (22% discount).

Two scoring modules

๐ŸŽ™๏ธ
Voice MOS: ViSQOL

Perceptual quality model developed by Google Research. Spectro-temporal analysis of the audio signal to produce an objective MOS score correlated with human subjective evaluations. Designed for VoLTE, VoNR and VoIP calls.

ViSQOL Google Research VoLTE MOS 1 to 5
EUR 4,490 /year
๐ŸŽฌ
Video MOS: ITU-T P.1204.3

ITU-T P.1204.3 parametric model for video quality. Evaluates MOS from stream metadata (bitrate, resolution, framerate, codec) without full video decoding. Ideal for real-time QoE monitoring on streaming platforms.

ITU-T P.1204.3 Parametric Streaming QoE Real-time
EUR 4,490 /year
Voice + Video Bundle

Both modules combined with a 22% discount.

EUR 6,990 /year EUR 8,980

Technical specifications

ARM64 native

Libraries compiled in C/C++ for ARM64. Real-time performance on Snapdragon: voice scoring under 50 ms per 3-second segment.

JNI Ready

JNI (Java Native Interface) ready for Android integration. Direct calls from Kotlin or Java, no additional wrapper needed.

Pre-trained models

Optimized ML models shipped ready to use. No client-side training required. Model updates included in the annual license.

VoLTE integration

Native correlation with HiCellTek radio KPIs: MOS vs RSRP, MOS vs SINR, MOS vs jitter. Root cause identification for quality degradation.

Who is it for?

Operator QoE teams

Measure perceived service quality for your subscribers on VoLTE/VoNR calls and video streaming. Identify degradation zones.

VoLTE integrators

Integrate MOS scoring into your VoLTE test solutions to validate end-to-end voice quality during deployments.

Streaming platforms

Real-time QoE monitoring for OTT video services. Proactive detection of video quality degradation.

Frequently asked questions

What is the difference between ViSQOL and ITU-T P.1204.3?

ViSQOL (Google Research) is a perceptual quality model for voice and audio based on spectro-temporal analysis. ITU-T P.1204.3 is a parametric model for video quality that uses stream metadata (bitrate, resolution, codec) without requiring full video decoding.

Does the MOS SDK work in real-time on Android?

Yes. Both models are compiled as native ARM64 with optimized pre-trained models. ViSQOL scoring processes a 3-second audio segment in under 50 ms on a Snapdragon 8 Gen 2.

Can I use the MOS SDK to measure VoLTE quality?

Yes. The ViSQOL voice module is designed to evaluate VoLTE and VoNR call quality. It produces a MOS (Mean Opinion Score) from 1 to 5 correlated with human subjective evaluations.

Does the voice + video bundle offer a discount?

Yes. The bundle combines both SDKs at EUR 6,990/year instead of EUR 8,980/year separately, a 22% discount.