|dc.description.abstract||Voice over Internet Protocol (VoIP) is an active area of research in the world of
communication. The high revenue made by the telecommunication companies is a
motivation to develop solutions that transmit voice over other media rather than
the traditional, circuit switching network.
However, while IP networks can carry data traffic very well due to their besteffort
nature, they are not designed to carry real-time applications such as voice.
As such several degradations can happen to the speech signal before it reaches its
destination. Therefore, it is important for legal, commercial, and technical reasons
to measure the quality of VoIP applications accurately and non-intrusively.
Several methods were proposed to measure the speech quality: some of these
methods are subjective, others are intrusive-based while others are non-intrusive.
One of the non-intrusive methods for measuring the speech quality is the E-model
standardised by the International Telecommunication Union-Telecommunication Standardisation
Although the E-model is a non-intrusive method for measuring the speech quality,
but it depends on the time-consuming, expensive and hard to conduct subjective
tests to calibrate its parameters, consequently it is applicable to a limited number
of conditions and speech coders. Also, it is less accurate than the intrusive methods
such as Perceptual Evaluation of Speech Quality (PESQ) because it does not consider
the contents of the received signal.
In this thesis an approach to extend the E-model based on PESQ is proposed.
Using this method the E-model can be extended to new network conditions and
applied to new speech coders without the need for the subjective tests. The modified
E-model calibrated using PESQ is compared with the E-model calibrated using
subjective tests to prove its effectiveness.
During the above extension the relation between quality estimation using the
E-model and PESQ is investigated and a correction formula is proposed to correct
the deviation in speech quality estimation.
Another extension to the E-model to improve its accuracy in comparison with
the PESQ looks into the content of the degraded signal and classifies packet loss
into either Voiced or Unvoiced based on the received surrounding packets. The accuracy
of the proposed method is evaluated by comparing the estimation of the new
method that takes packet class into consideration with the measurement provided
by PESQ as a more accurate, intrusive method for measuring the speech quality.
The above two extensions for quality estimation of the E-model are combined
to offer a method for estimating the quality of VoIP applications accurately, nonintrusively
without the need for the time-consuming, expensive, and hard to conduct
Finally, the applicability of the E-model or the modified E-model in measuring
the quality of services in Service Oriented Computing (SOC) is illustrated.||en