| The dissertation takes Nanjing dialect intonation as research object,through acoustic experiments,investigates the realization of stress and the functional tones of Nanjing dialect,including statement,question,exclamatory,and imperative tones in the context of speaking.The conclusions of this dissertation are as follows:Two-character tone research.(1)The word stress of the two-character tone is mainly achieved through pitch elevation,range expansion and tone enhancement.The effect of duration is not prominent,but the listening recognition of tonal stress of two-character with the first word in Yangping and Entering depends on duration.There is a relatively consistent correspondence between the pitch,range,and intensity of each tone.Tones with a higher pitch curve and a wider range are also stronger.(2)The acoustic performance of the two-character tone at the beginning and the middle of the sentence is relatively close,which is more affected by the intonation of the statement.The duration,pitch,and intensity of the different tones of the preceding and following characters are less distinguished,and their boundaries are blurred,while there is a strong regularity in the distribution of the duration of the preceding and following words.The two-character groups with Yinping,Yangping and Qusheng as the proceeding character have longer former word,while the group with Shangsheng and Rusheng have longer back word.The types of tone combinations in sentence are generally less than those in monophonic mode.The former words in high flat tone(44)and midlle flat tone(33)are accounted for the largest proportions,while the back words in high flat tone(44)and high falling tone(43)are accounted for the largest proportion.(3)At the end of the sentence,the two-character tones have clear boundaries,whose characteristics are similar to those of monophonic,but their pitch scale is significantly reduced.(4)The pitch of two-character tones are not significantly expanded or narrowed in statement sentences compared to single-character tones.Stress research.(1)The emphasizing stress in Nanjing dialect is mainly realized by the highlighting of duration in sentences,and the narrowing of the pitch range and elevation of the scale caused by the raising of the lowest limit of pitch.The propotion of average duration of accented syllables in a sentence reaches 19.7%,which shows a significant duration increase in the sense of hearing.The pitch range is narrower than the tone of single character,and the tone intensity has little effect on the realization of accent emphasis.(2)Stress at the front of the sentence(subject,predicate,and object)forms a prosodic unit by itself,which is similar in characteristics to narrow focus.Its scale is higher than the unstressed syllable,and its lowest limit of pitch is significantly higher than that of a single character and the adjacent syllables in the sentence.Its pitch range may not be prominent,but it will cause the range of the syllables before and after to expand or that their highest and lowest limit will rise.(3)Stress at the back of the sentence(object complement)won't form a prosodic unit on its own,whose characteristics are similar to the broad focus.Its pitch scale and range are not prominent,but each character of the prosodic words where the stress is located often have scale ups.(4)The fundamental frequency curve and the trend of the pitch of the accented sentence are in line with the downward pitch tendency,but if the sentence has two or more prosodic words,the prosodic word will form two or more pitch peaks at the stress.(5)When the stress falls in the end of the sentence or the end of prosodic words,the average range of the sentence is larger.(6)The low point of the boundary tone of the accented statement sentence is lower than that of the preceding character,and the high point is basically the same as or slightly higher than its preceding character,which causes the range of the boundary tones to be expanded compared to the preceding character.Functional tone research.(1)Question sentences in Nanjing dialect are manifested as a significant increase in the entire sentence pitch scale.The range of the syllable in the middle of the sentence is larger than that of the statement sentence,but the range of the first character and boundary tones is not always larger than that of the statement sentence,and the range expansion is mainly achieved through the elevation of the highest limit of the pitch.The tone intensity of the first word and boundary is enhanced,and their duration is shortened.(2)The boundary tone range of question sentences is significantly larger than that of the first character,and its reverse synergistic effect on adjacent syllables is more obvious than the forward synergistic effect of the first character to adjacent syllables,and it carries more question information.(3)Exclamatory sentences in Nanjing dialect are generally realized by raising the pitch scale of the whole sentence,raising the lower limit of pitch,highlighting the boundary duration,strengthening the tone intensity of the boundary tone and the first tone.(4)Stress has a significant impact on the realization of exclamatory sentences.The duration of the accented position is generally greater than the first word and the boundary.Stress is stronger at the beginning and end of an exclamation sentence,while in the sentence it is manifested as a reduction in the range,a rise in low pitches,and an increase in the intensity of the sound.(5)The pitch scales of the front-accented and middle-accented exclamation sentences are not significantly elevated compared to the statement sentences.The boundary tone duration of the middle-accented exclamation sentences is still longer than that of the first character.The realization of the back-accented exclamation sentences is similar to that of the full-accented sentence,and the pitch scale of sentence has a general raise of two scales,the first tone is high and flat,and the boundary tone is medium and flat.On the whole,the performance of the back-accented exclamation is similar to that of the full-accented,and the front-accented and middle-accented are similar.(6)The main way to realize imperative sentences in Nanjing dialect is to shorten the duration,raise the pitch of the whole sentence,narrow the pitch range,and increase the intensity of the whole sentence significantly.PENTAtrainer intonation synthesis research.(1)The parameter training effect of multi-person statement sentence with emphasized accent in PENTAtrainer is good,with large mean RSME,but good correlation performance,and the naturalness and recognition rate of synthesized speech are high.(2)When the stress is in the position of the object complement(noun)at the end of the sentence,the mean RSME of the q TA parameter is the smallest;when the stress is in the position of the object complement verb and quantifier in the sentence,the mean RSME is the largest,the pitch performance is unstable,and the correlation is lower,the synthesized speech is the worst.(3)When the accented syllables are Yinping and Entering,the naturalness of the synthesized speech is the highest,and the recognition rate reaches 100%.When the accented syllable is Shangsheng,the recognition rate of the synthetic speech accent is the lowest.(4)The parameter training results of the multi-person multi-functional tone corpus are the worst,while the single-person single-function tone corpus has the best effect,whose mean RSME is small and average correlation is greater than 0.8,and the naturalness and tone recognition rate are high.(5)The order of parameter training effect from good to bad is: statement sentence> question sentence> exclamatory sentence,and the MOS score and recognition rate from high to low are sorted as: statement sentence> exclamation sentence> qusetionsentence.The overall effect of the parameter training of each functional tone of male speakers is better than female speakers.(6)As for exclamatory sentences and question sentences obtained by modifying the pitch target parameters of statement sentences,the average recognition rate is below 80%,and the effect is not ideal.The speech recognition rate of question sentences after parameter modification is lower than that of exclamatory sentences. |
PDF Full Text Request