| Background and Objective:Objective acoustic speech analysis is a valuable technique that enables speech clinicians to compare speech and distinguish it as normal and abnormal.The purpose of this study is to screen out more parameters for the diagnostic analysis of voice.At the same time,the voices of patients with T1 glottic cancer,unilateral vocal cord paralysis,vocal cord leukoplecta and vocal cord polyp were respectively verified to see whether there were differences between them and normal people.In addition,the accuracy of Chinese vowels/a/,/o/,/e/,/i/,/u/and/u/in the diagnosis of vocal cord diseases between patients and normal people was verified.Method:In this study,30 normal healthy adults in the control group and 134 patients with vocal cord diseases(28 subjects in the glottic cancer group,32 subjects in the vocal cord paralysis group,30 subjects in the vocal cord leukoplachia group,and 44 subjects in the vocal cord polyp group)were collected and their speech samples of six Chinese vowel sounds were collected.A total of 33 acoustic speech parameters were extracted.The acoustic parameters of the four disease groups were compared with those of the control group.Results:1.When the patients with vocal cord paralysis pronounce Chinese vowel/a/,there was significant difference in parameter last_f0,hnr,duration between the group and the normal group(p<0.05).The vowel/o/in duration was significantly different from that of normal people(p<0.05).When Chinese vowel/e/was pronounced,there was a significant difference between duration-hr_median and normal people(p<0.05).When Chinese vowel/i/was pronounced,there was a significant difference in skew between normal and Chinese vowel(p<0.05).In the Chinese vowel/u/,there were significant differences between normal people and middle_f0,hnr,duration,hr_mean and hr_std(p<0.05).When Chinese vowel/u/was pronounced,there were significant differences between the normal people and the skew,duration-hr_median(p<0.05).The area under ROC curve(AUC)of vocal cord paralysis group was all greater than 0.5,indicating that the prediction model had a certain accuracy.The vowel/i/has the largest area under the curve and its AUC=0.7026.2.There were significant differences between vocal cord polyp group and normal people in terms of parameters first_f0,last_f0,slope_max2end,duration and hr_min when Chinese vowel/a/(p<0.05).In the Chinese vowel/o/,there were significant differences with normal people in parameters f0variation,range_f0,duration,hr_max and hr_min(p<0.05).When Chinese vowel/e/was expressed,it had significant differences with normal people in parameters f0variation,max_f0,range_f0,duration and hr_max(p<0.05).There were significant differences in the parameters std,rms,energy and duration between vowel/i/and normal people(p<0.05).There were significant differences with normal people in parameters f0variation,f0skew,max_f0,range_f0,slope_max2end-duration and hr_min when Chinese vowel/u/was released(p<0.05).The vowel/u/in duration was significantly different from that of normal people(p<0.05).The area under ROC curve(AUC)of vocal cord polyp group was all greater than 0.5,indicating that the prediction model had a certain accuracy.Vowel/u/has the largest area under the curve and its AUC=0.6752.3.There were significant differences between patients in vocal cord leukoplecia group and normal people in the parameters mean_f0 and duration when they pronounced Chinese vowel/a/(p<0.05).There were significant differences between the Chinese vowel/o/and normal people in the parameter hr_std(p<0.05).When Chinese vowel/e/was pronounced,there were significant differences between them and normal people in parameters first_f0,median_f0,mean_f0,duration and hr_std(p<0.05).There were significant differences in the parameters skew and hr_std between Chinese vowel/i/and normal people(p<0.05).There were significant differences in the parameters hr_std and hr_max between Chinese vowel/u/and normal people(p<0.05).There were significant differences between Chinese vowel/u/and normal people in skew and hr_std(p<0.05).The area under ROC curve(AUC)of vocal cord leukoplakia group was all greater than 0.5,indicating that the model prediction had a certain accuracy.Vowel/u/has the largest area under the curve and its AUC=0.6897.4.There were significant differences between glottic cancer group and normal people in parameters skew,f0variation,max_f0,range_f0 and duration when Chinese vowel/a/was pronounced(p<0.05).There were significant differences between the Chinese vowel/o/and normal people in the parameters skew and f0variation(p<0.05).mean_f0,f0skew,slope_start2max,during-hr_std were significantly different from normal people when Chinese vowel/e/was pronounced(p<0.05).There was a significant difference between Chinese vowel/i/and normal people in hr_std(p<0.05).The parameters median_f0,mean_f0,f0skew,max_f0,range_f0,slope_max2end,duron-hr_median and hr_min when Chinese vowel/u/were pronounced were significantly different from those of normal people(p<0.05).There were significant differences between Chinese vowel/u/and normal people in skew and hr_std(p<0.05).The area under ROC curve(AUC)of vocal cord cancer group was all greater than 0.5,indicating that the prediction model had a certain accuracy.Vowel/o/has the largest area under the curve and its AUC=0.731.Conclusion:T1 glottic cancer,unilateral vocal cord paralysis,vocal cord leukoplakia,vocal cord polyp patients compared with normal human voice differences.A number of statistically significant acoustic parameters were also identified that may distinguish the diagnosis of each of the four vocal cord disorders from that of normal people.But whether they have clinical significance will need to be further tested by obtaining more samples of subjects with vocal cord disease at various levels of dysphonia.In addition,the Chinese vowel/o/had the highest accuracy in distinguishing T1 glottic cancer patients from normal people.The Chinese vowel/i/had the highest accuracy in distinguishing patients with unilateral vocal cord paralysis from normal ones.In addition,Chinese vowel/u/had the highest accuracy in distinguishing patients with vocal cord leukoplusand polyps from normal people.These results need to be further validated clinically using voice samples from more subjects. |
PDF Full Text Request