Quality of service (QoS) analysis frameworkn for text to speech (TTS) services

• Main Author: Mohamed Hassan, Mohamud
• Format: Thesis
• Language: English; English; English
• Published: 2015
• Subjects: TK7800-8360 Electronics
• Online Access: http://eprints.uthm.edu.my/1357/2/MOHAMUD%20MOHAMED%20HASSAN%20COPYRIGHT%20DECLARATION.pdf; http://eprints.uthm.edu.my/1357/1/24p%20MOHAMUD%20MOHAMED%20HASSAN.pdf; http://eprints.uthm.edu.my/1357/3/MOHAMUD%20MOHAMED%20HASSAN%20WATERMARK.pdf; http://eprints.uthm.edu.my/1357/

Quality of service (QoS) evaluations is significant and necessary for text to speech web service applications. Text to speech media conversion quality measurements has general and specific mechanisms for its functional and nonfunctional requirements. The main objective of this thesis is to introduce QoS framework which is able to evaluate and analyze the perceived quality of services (QoS) for text to speech (TTS) web services. To achieve this goal, the framework combines two main mechanisms for measuring the speech quality. General quality attributes measure the response time of TTS services, specific quality attributes measure intelligibility and naturalness through subjective quality measurements, which are mapped onto mean opinion score (MOS). Twenty individuals participated the experiment to test the speech quality by comparing three services fromtexttospeech.com, Natural Reader and Yakitome. Aggregate scores has been used to calculate the combination of general and specific nonfunctional QoS on TTS Web services. The result shown better scale for quality estimation, service1 (Fromtexttospeech) 47.84% is suitable TTS service provider where service2 and service3 (NaturalReader and Yakitome) are close 31.62 and 21.53% respectively and less preferred for listening tests to assess synthesized speech. It is essential to consider the user’s perspective when evaluating the quality of services for media conversion services such as text to speech (TTS) to enhance the user experience.