A Study Of The Effect Of Gender On The Production And Perception Of Mandarin Lombard Speech

Daily communication is disturbed by various kinds of noise.Previous studies had shown that noisy environments can induce speech modifications,which is known as Lombard speech(Lombard,1911).Although studies of Lombard speech have been carried out in western countries for quite a long time,however,most studies of Lombard speech have focused on non-tonal languages,such as English,French,Dutch,while few studies have investigated the acoustic features of tonal languages produced in noise,such as Mandarin.Additionally,previous studies have indicated significant gender differences in the production of plain speech(speaking in quiet),but few studies have discussed the gender difference in the speech production in noise.Furthermore,for the studies of speech perception in noise for non-native listeners,the experimental materials are mostly limited to plain speech,while previous researches have suggested that Lombard speech is more intelligible than plain speech in noise because of its high pitch and low speed rate.However,it is unclear whether the intelligibility of Lombard speech can contribute to non-native listeners in noise.Inspired by the above research shortcomings,two experiments(speech production and speech perception)were conducted in this study to explore the gender difference in acoustic realization of Mandarin in noise and the perception of Mandarin Lombard speech in noise for native and non-native listeners.More specifically,the study tried to answer the following three research questions:1)What is the gender difference in acoustic features of Mandarin produced in noise?2)Is Lombard speech more intelligible than plain speech in noise for native listeners?3)Is Lombard speech more intelligible than plain speech in noise for non-native listeners?A total of 20 native Mandarin speakers were recruited in the production experiment,this experiment recorded their speech production in both quiet and noise conditions.After recording,various acoustic features of the corpus were analyzed,such as duration,intensity,pitch,vowel formant frequencies,tone space and vowel space,then the speech modifications made by speakers in noise condition were explored.A total of 30 native Mandarin listeners and 11 non-native listeners were recruited to participate in the perception experiment.The materials used in the perception experiment were taken from the corpus recorded by six speakers in the production experiment,then they were mixed with speech-shaped noise and adjusted to different signal-to-noise ratios(-9,-12,-15 d B).The participants need to identify the materials played in the headphones and make choice on the computer screen.The accuracy of listeners’ identification in noise was analyzed to investigate the intelligibility of Lombard speech in the perception experiment.The main conclusions of this thesis are as follows:1)In Lombard speech,the tonal duration is significantly longer in females than in males.In addition,pitch and intensity are significantly higher in females than in males in both Lombard and plain speech,but no significant gender differences are observed in pitch slope and tone space.There is a significant increase in vowel formant frequencies in Lombard speech,and the hyper-articulation in F1 is more pronounced than in F2.Nevertheless,no significant expansion of vowel space is found in Lombard speech;2)For native listeners,the accuracy of tone identification in Lombard speech in noise is significantly higher than plain speech.Thus,intelligibility of Lombard speech in noise is higher than that of the plain speech,especially the female speech;3)Non-native listeners also perceive Lombard speech more correctly in noise than plain speech.In addition,non-native listeners perceive the speech of female speakers better than males,suggesting that the intelligibility of female Lombard speech in noise also contributes to non-native listeners.In summary,Lombard speech in noise modulated in several acoustic parameters,such as duration,intensity,and pitch,but no significant expansion of tone space and vowel space was found.Furthermore,the changes in these acoustic parameters would be beneficial to improve the intelligibility of Lombard speech in noise,especially for female speech.