The automatic detection of heart failure using speech signals

• For the first time, automatic detection of heart failure (HF) from speech is studied. Both vocal tract and glottal source parameters are used in feature extraction. Four machine learning algorithms as classifiers. Applying Feature selection on + features improved classification accuracy. Among classifiers, neural network gave best performance. Heart a major global health concern increasing prevalence. It affects larynx breathing – thereby quality speech. In this article, we propose an approach for people with HF using signal. The proposed method explores mel-frequency cepstral coefficient (MFCC) features, their combination to distinguish healthy were extracted voice signal estimated inverse filtering. algorithms, namely, support vector machine, Extra Tree, AdaBoost, feed-forward (FFNN), trained separately individual combination. was observed that MFCC yielded higher accuracies compared features. Furthermore, complementary nature investigated by combining these Our results show FFNN classifier reduced set achieved overall performance both speaker-dependent speaker-independent scenarios.