Mean opinion score (MOS) revisited: methods and applications, limitations and alternatives

By Robert C. Streijl, Stefan Winkler, David S. Hands. 2016

520 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Tao Qin
2. Tie-Yan Liu
3. Luc Van Gool
4. Xu Tan
5. Ming-Hsuan Yang
6. Ferenc Huszár
7. Dacheng Tao
8. Xiaogang Wang

Citations in the corpus

• Semi-Supervised Neural Architecture Search (2020, methodology): “We also use MOS (mean opinion score) [28] to evaluate the naturalness of the synthesized speech.”
• Lossy Image Compression with Compressive Autoencoders (2017, methodology): “While MOS tests have their limitations, they are a widely used standard for evaluating perceptual quality (Streijl et al., 2014).”
• Improving transfer of expressivity for end-to-end multispeaker text-to-speech synthesis (2021, methodology): “Likewise, for expressive MOS, listeners are directed to provide scores between 1 (bad) and 5 (excellent) depending on how synthesized speech utterance resembles the expressivity given in the reference speech utterance.”
• On the use of deep learning for blind image quality assessment (2016, methodology): “The CNN is discriminatively fine-tuned to classify image sub-regions into five classes according to the 5-points MOS scale [43].”
• Open-source Multi-speaker Speech Corpora for Building Gujarati, Kannada, Malayalam, Marathi, Tamil and Telugu Speech Synthesis Systems (2020, methodology): “The voice was evaluated by the raters using a Mean Opinion Score (MOS) listening test (Streijl et al., 2016).”
• Neural Text-to-Speech Adaptation from Low Quality Public Recordings (2019, background): “Synthesized voices are evaluated from three aspects: 1) the speech naturalness Mean Opinion Score (MOS) (from 1 to 5) [34] 2) speaker similarity MOS (from 1 to 5) compared to the target speaker voice 3) Signal Noise Ratio (SNR).”
• Incremental Text to Speech for Neural Sequence-to-Sequence Models using Reinforcement Learning (2020, methodology): “There are two aspects of the agent’s performance that we track: Quality: We compute the Mean Opinion Score (MOS) to measure the naturalness of our audio [37, 4].”
• An Overview of Voice Conversion and Its Challenges: From Statistical Modeling to Deep Learning (2020, background): “(MOS) [2], [251]–[253], preference tests [18], [254] and bestworst scaling [255] could represent the intrinsic naturalness and similarity to the target.”