Research And Implementation Of Emotional Speech Synthesis Based On Association Rules And Waveform Concatenation Combination

With the constant development of modern computer technology, human-computer interaction plays an increasingly important role in people’s lives. Voice communication between people and computers communicate the most direct and convenient means of communication, it is first and foremost involved in speech synthesis technology. The conventional speech synthesis technology is focused on the accuracy of the speech word expression, lack of voice emotion, synthesized speech having a strong flavor of the plant. How to make your computer synthesized voice as clear as people speak intelligibility, high natural and full of rich emotional speech synthesis workers a long-term and arduous task.Firstly emotional speech synthesis development and technical research, reported on the status of the emotional speech synthesis technology, compared the characteristics of each of the mainstream speech synthesis technology, combined with the emotion of the voice analysis, the use of data mining association rules splicing techniques and waveform synthesis of natural high clear intelligibility emotional voice. The thesis abstract a more intuitive class sinusoidal fundamental frequency curve model, given the fundamental frequency of association rules model, Apriori algorithm of association rules has been improved, drawn before and after the syllable sound associated, thus to guide the selection of an emotional speech synthesis sound, narrowing the scope of the waveform concatenation, improve waveform stitching effect, and the use of TD-psola algorithm prosody parameters were adjusted.The experimental corpus from the library of the world’s most professional voice software vendors Loquendo speech technology research laboratory Dave’s voice.Secondly, in the build system module development process, the paper proposes emotional speech synthesis techniques and algorithms based on emotional speech synthesis system is designed and implemented, and synthesized emotional voice was angry, happy, sad and fear comparative analysis of the kinds of emotions and emotional features of the human voice. The system is easy to operate, user-friendly, high stability. The results show that emotional speech synthesis using the system characteristic parameters is very close to the characteristic parameters of the real emotional statement, the synthesized voice in the hearing on the high natural, clear and understandable at the same time full of rich emotional. In other words, a combination of research methods based on association rules and waveform concatenation is an ideal. The emotional speech synthesis provides help and guidance.