A Study Of The Tone Of Chinese Vowels Recognition Based On Spectrogram

Speech emotion recognition is not only an important part of speech recognition,butalso an important research direction of realizing human-computer interaction.Chinese is atonal language;its tone changes can express semantic information and reflect the speaker’semotions. Chinese tones are divided into four types(except the untoned),the1st tone(high-level tone),2nd tone(rising tone),3rd tone(falling-rising tone) and4th tone(high-fallingtone).The variation is mainly reflected in the vowel changes. This article is mainly targetedat the four tones of Chinese basic vowels(single vowel).In this paper,the study summarizes the common Chinese tone recognitionalgorithm.On the basis of this,the author proposes another algorithm about tone recognitionof Chinese pure vowels.It does not adopt the method of fundamental frequency detection,but applies the mathematical morphology image processing techniques to analyse thefeatures of Chinese vowels’tone,based on its holistic characteristics of the spectrogram.Atthe same time,mathematical morphology image processing techniques will be applied tothe field of speech recognition.This study uses MATLAB7.1software as the algorithm to complete the programmingand simulation.Firstly,it is necessary to record the original speech signals,and then,theauthor converts it into spectrogram after preprocessing with the Cooledit Pro2.0software.Secondly,the spectrogram is pre-processed with image processing technology andmathematical morphology image processing technology,in order to further complete to theimage analysis and the extraction of their features,such as the operation ofsmoothness,normalization,and so on.Next,the morphological characteristics of everyspectrogram are observed and analysed.The author extracts the morphological features ofthe spectrogram by using skeleton operation.And then, the results after skeletonization areaccounted for in each direction.Finally, the main directions of four tones go through T testby using SPSS17.0software to determine whether there are significant differences.It isdetermined that whether the result is effective.