Dissertation > Excellent graduate degree dissertation topics show

The Pre-research in Application of Cough Monitor

Author: MaZuo
Tutor: ZhengZeGuang
School: Guangzhou Medical College
Course: Respiration within the science
Keywords: Voice recognition technology Hidden Markov Models Chronic cough Cough objective record
CLC: R725.6
Type: Master's thesis
Year: 2011
Downloads: 19
Quote: 0
Read: Download Dissertation

Abstract

The first part of the cough Automatic Identification System software initially established purpose: cough use voice recognition technology to create automatic identification system software for the automatic recognition of cough sound. Methods: 46 Guangzhou Institute of Respiratory Diseases inpatients as subjects. Subjects in a quiet environment one hour recording microphone affixed to the inner edge of the patient sternocleidomastoid speech check subjects AC 10 minutes, so that can be recorded at the same time to the speech and coughing sound. If the patient does not cough chili powder can be used to induce the disease cough. Recording equipment using the the Dell original machine sound card through the sampling rate of 8kHz, quantify median 16bit, single-channel A / D conversion is converted into a digital signal is stored in the recording device. We recording samples of 46 subjects were divided into two groups, the first group of 34 subjects recording samples, as a hidden Markov model training sample, extraction of training samples Mel Frequency Cepstral Coefficients (MFCC) establish hidden Markov model, combined endpoint detection technology written cough Automatic Identification System software; recording samples of the second group of 12 subjects as test samples, and the use of automatic identification system software identification cough signals according to the above written cough. Cough recording samples of 46 subjects were artificial recognition results as the gold standard to compare results obtained in accordance with the the artificial identification and system software to identify and calculate: the missed endpoint detection technology = undetected fragment of cough / actual total cough fragment; insertion rate = the voice fragment mistakenly identified as cough / suspicious cough fragments; to cough, automatic identification system sensitivity = correctly identified as the the cough cough fragments / actual total cough fragment; specificity = correctly recognized as the voice of the voice clips / voice identified as the fragment. Results: the artificial recognition as a training sample of 34 subjects recording sample obtained 556 cough sample 1832 and interference samples, cough sample of long-time classification of cough and interference template, respectively, using the corresponding sample data on cough and cough template training 5 the cough template and a filler template (fill in blank template: voice training by coughing outside income template). Cough prepared Automatic Identification System for automatic identification, as test samples, a sample of 12 subjects the recording will automatically identify the results compared to the artificial recognition result obtained: undetected rate and the rate of the insertion of the endpoint detection techniques. 3.09% and 34.19%; cough automatic identification system the software identification cough sound sensitivity and specificity were 96% and 86%. Automatic Identification System software Conclusion: The use of speech recognition technology to create the cough cough sound can be used to automatically identify, provide the premise for the development of cough recording analyzer. Analysis of the purpose of the second part of the characteristics of different causes of chronic cough: The the different chronic cough cough analysis to understand the different causes of chronic cough cough characteristics, exploring the use of cough logger the automatic identification cough to make chronic cough, and according to their coughing features diagnosis. Methods: Between September 2009 to January 2011, during the chronic cough clinic at the Guangzhou Institute of Respiratory Diseases outpatient or ward patients, recording 24 hours of patients with chronic cough cough sound, and at the same time teach patients learn to record with cough related events : include into to get up, eat, sleep. To give checks in accordance with the characteristics of the patient's history, follow the guide: pulmonary ventilatory function in bronchial provocation test, the sputum induction cytology and 24-hour esophageal PH value monitoring. Recorded 24-hour cough signals manual analysis, labeled: dry cough, wet cough, mono cough, alliteration cough, the continuity cough, clear throaty, Qing sputum tone and bronchospasm tone. Cough number of different time periods according to the different causes of group (to get up, eat, sleep) for statistical analysis. The SPSS16.0 systems analysis, multivariate logistic regression analysis to arrive at a final conclusion; within the group and between groups were compared using the nonparametric rank sum test. Results: A total of 100 cases were collected, has completed the analysis of 82 cases of patients with chronic cough. The the cough characteristics and regularity of patients with CVA: the total cough the number of the sleep period total cough / 24 hours = 27.81%, 4, 5 and 6 hours of sleep period cough / sleep period total cough = 45%, dry cough apparent more than wet cough (P lt; 0.05), more associated with bronchospasm (P lt; 0.05). Is CVA variables, parameters (the dry cough X ₁ wet cough X ₂ sleep period total cough number X ₃, bronchial the spasm tone X ₄ Qing guttural and phlegm tone X5 morning one hour the the cough number of X ₆, before going to sleep the 1 hour cough number X _{7 < / sub>) as independent variables, multivariate logistic regression analysis. Draw probability equation: P = e ^{4.34X ₃ 5.49X ₄ -3.16 / 1 e ^{4.34X ₃-5.49X ₄ -3.16 (P lt; 0.001) UACS coughing characteristics and regularity of the patients: early morning coughing significantly (P lt; 0.05) to wet cough main (P lt; 0.05), many accompanied by clear the sputum tone and Qing guttural (P lt; 0.05). As the dependent variable is UACS parameters (the dry cough number X ₁, the number of wet cough X ₂ Qing guttural phlegm tone X ₃, the morning after one hour the the cough number of X ₅, one hour before going to sleep cough number X6) as independent variables, multivariate logistic regression analysis. Draw probability equation: P = e ^{2.16X ₂ 2.72X ₃ -6.17 / 1 e ^{2.16X ₂ in 2.72X -6.17 (P lt; 0.001) the GERC the patients cough characteristics and regularity as: cough obvious after dinner, before meals 1,2 hours cough number and 1 and 2 hours after a meal to cough Variance P lt; 0.05; mainly dry cough, lt of p wet cough; 0.05; whether GERC as the dependent variable, the parameters (dry cough X _X _{2 meal 1 hour X ₃ 2 hours after meals X ₄, one hour after dinner X}, moist cough the ₅, two hours after dinner X ₆) as independent variables in multivariate logistic regression analysis. Draw probability equation: P = e ^{0.91X ₃ 1.84X ₅ -7.99 / 1 e ^{0.91X ₃ in 1.84X -7.99 (P lt; 0.001) EB cough frequency no apparent regularity, mainly dry cough, wet cough P-lt; 0.05; due two cases, a small amount of nighttime cough, so the main differential diagnosis and CVA. Conclusion: After a series of related analysis, can be drawn from the different causes of chronic cough regression probability equation, and a high clinical value, you can create a simple diagnostic method for chronic cough and clinical application. Summary of the study: a two-part study, not only the initial establishment cough automatic identification system, also a simple diagnostic method for the diagnosis of chronic cough for 24 hours cough recorded analyzer developed provides a study of the feasibility and methodology .}}}}}}}

Related Dissertations

CLC: > Medicine, health > Pediatrics > Children within the science > Department of pediatric respiratory and chest diseases
© 2012 www.DissertationTopic.Net Mobile