(1) Real-Time, Non-Intrusive Evaluation of VoIP (2) (i) Muhammad Adil Raja, F1-OP-08 Foundation Building University of of Limerick Ireland adil.raja@ul.ie +353-61-202715 (ii) Raja Muhammad Atif Azad CSIS University of Limerick Ireland atif.azad@ul.ie +353-61-202763 (iii) Colin Flanagan F2-0-04 Foundation Building University of Limerick Ireland colin.flanagan@ul.ie +353-61-202622 (iv) Conor Ryan CSIS University of Limerick Ireland conor.ryan@ul.ie +353-61-202755 (3) Muhammad Adil Raja (4) Speech quality, as perceived by the users of Voice over Internet Protocol (VoIP) telephony, is critically important to the uptake of this service. VoIP quality can be degraded by network layer problems (delay, jitter, packet loss). This paper presents a method for real-time, non-intrusive speech quality estimation for VoIP that emulates the subjective listening quality measures based on Mean Opinion Scores (MOS). MOS provide the numerical indication of perceived quality of speech. We employ a Genetic Programming based symbolic regression approach to derive a speech quality estimation model. Our results compare favorably with the International Telecommunications Union-Telecommunication Standardization (ITU-T) PESQ algorithm which is the most widely accepted standard for speech quality estimation. Moreover, our model is suitable for real-time speech quality estimation of VoIP while PESQ is not. The performance of the proposed model was also compared to the new ITU-T recommendation P.563 for non-intrusive speech quality estimation and an improved performance was observed. (5) B, D, E, G (6) Our results are better than past research. Specifically we have presented a scheme to quantify the significance of each of the impairments of a VoIP system. Our scheme also prunes off a number of redundant parameters and presents befitting models for quality estimation. The models are also unique and counter-intuitive. Thus we solve a long-held formidable problem using GP. Our models approximate the ITU-T PESQ algorithm fairly accurately, where latter is an industry wide "intrusive" algorithm for speech quality estimation. PESQ is the most reliable, but compute-intensive, algorithm but uses a reference model to perform DSP based estimation of speech quality; our scheme is non-intrusive and real-time. Our model also outperforms, by a considerable margin, the ITU-T Recommendation P.563 which is the current standard for non-intrusive speech quality evaluation. Thus, The evolved models combine the best of both the worlds (i.e. intrusive and non-intrusive) by demonstrating a strong correlation with PESQ and being very cheap to execute because: - they are very parsimonious mathematical expressions. - they involve a surprisingly small number of input variables. (7) @InProceedings{eurogp07:raja, author = "Adil Raja and R. Muhammad Atif Azad and Colin Flanagan and Conor Ryan", title = "Real-Time, Non-Intrusive Evaluation of Vo{IP}", editor = "Marc Ebner and Michael O'Neill and Anik\'o Ek\'art and Leonardo Vanneschi and Anna Isabel Esparcia-Alc\'azar", booktitle = "Proceedings of the 10th European Conference on Genetic Programming", publisher = "Springer", series = "Lecture Notes in Computer Science", volume = "4445", year = "2007", address = "Valencia, Spain", month = "11 - 13 " # apr, keywords = "genetic algorithms, genetic programming", pages = "217-228", notes = "Part of \cite{ebner:2007:GP} EuroGP'2007 held in conjunction with EvoCOP2007, EvoBIO2007 and EvoWorkshops2007", } (8) To be divided equally among the authors. (9) Our work on speech quality estimation is seminal and clearly demonstrates that the evolved models approximate the ITU-T Recommendation P.862 (PESQ) fairly accurately and outperform ITU-T Recommendation P.563. Moreover, our models are also real-time efficient whereas both the aforementioned standards are not.