Respiratory Function Classification Algorithm Based On Cough Sounds And Software Development

Respiratory failure is the lending cause of death among patients with cervical spinal cord injury(CSCI).However,the traditional respiratory function detection methods are complicated and expensive,which brings great burden to doctors and patients.As a common symptom of respiratory system,cough can provide important information for judging the respiratory function of patients.Therefore,the assessment of respiratory function of patients with CSCI based on cough sound has important social value and application prospect.The goal of this paper is to judge the respiratory function of patients with CSCI based on cough sounds,and then develop a mobile phone client software to assist doctors in rapid analysis and diagnosis.The main contents include:1)Research and implementation of true and false cough recognition algorithm: A true cough sound is similar to an unvoiced sound,and a false cough sound similar to a voiced sound,in phonation mechanism.Using unvoiced signals and voiced signals as training samples,choosing zero-crossing rate and minimum amplitude difference coefficient as features,excluding the false cough of shouting type that easily lead to misjudgment respiratory function of patients with CSCI based on Fisher’s linear discriminant method.2)Research and implementation of respiratory function classification algorithm: By analyzing the characteristics of cough sounds in patients with CSCI,choosing Mel cepstrum coefficients and sparseness quantitative indicators of logarithmic spectrum as features,using support vector machine to build a classification model of the respiratory function of patients with CSCI.3)Implementation of mobile client software: Develop a mobile phone client software to assist doctors in rapidly analyzing and diagnosing the respiratory function of patients with CSCI based on the Android platform,including cough sound collection,waveform display,playback,true and false cough recognition,respiratory function evaluation and user information management,etc.In this paper,the validity of the true and false cough recognition algorithm and the respiratory function classification algorithm is tested through experiments;and the implemented mobile client software basically met the functional requirements and achieve the expected goal.