Heart sounds are critical for early detection of cardiovascular diseases, yet existing studies mostly focus on traditional signal segmentation, feature extraction, and shallow classifiers, which often fail to sufficiently capture the dynamic and nonlinear characteristics of heart sounds, limit recognition of complex heart sound patterns, and are sensitive to data imbalance, resulting in poor classification performance. To address these limitations, this study proposes a novel heart sound classification method that integrates improved Mel-frequency cepstral coefficients (MFCC) for feature extraction with a convolutional neural network (CNN) and a deep Transformer model. In the preprocessing stage, a Butterworth filter is applied for denoising, and continuous heart sound signals are directly processed without segmenting the cardiac cycles, allowing the improved MFCC features to better capture dynamic characteristics. These features are then fed into a CNN for feature learning, followed by global average pooling (GAP) to reduce model complexity and mitigate overfitting. Lastly, a deep Transformer module is employed to further extract and fuse features, completing the heart sound classification. To handle data imbalance, the model uses focal loss as the objective function. Experiments on two public datasets demonstrate that the proposed method performs effectively in both binary and multi-class classification tasks. This approach enables efficient classification of continuous heart sound signals, provides a reference methodology for future heart sound research for disease classification, and supports the development of wearable devices and home monitoring systems.
Heart Sounds/physiology* ; Humans ; Algorithms ; Neural Networks, Computer ; Signal Processing, Computer-Assisted ; Deep Learning ; Cardiovascular Diseases/diagnosis* ; Classification Algorithms

Fetal heart sound is nonlinear and non-stationary, which contains a lot of noise when it is colleced, so the denoising method is important. We proposed a new denoising method in our study. Firstly, we chose the preprocessing of low-pass filter with a cutoff frequency of 200 Hz and the resampling. Secondly, we decomposed the signal based on empirical mode decomposition method (EMD) of Hilbert-Huang transform, then denoised some selected target components with wavelet soft threshold adaptive noise cancellation algorithm. Finally we got the clean fetal heart sound by combining the target components. In the EMD, we used a mask signal to eliminate the mode mixing problem, used mirroring extension method to eliminate the end effect, and referenced the stopping rule from the research of Rilling. This method eliminated the baseline drift and noise at once. To compare with wavelet transform (WT), mathematical morphology (MM) and the Fourier transform (FT), the SNR was improved obviously, and the RMSE was the minimum, which could satisfy the need of the practical application.
Algorithms ; Fetal Heart ; physiology ; Heart Sounds ; Humans ; Signal Processing, Computer-Assisted ; Signal-To-Noise Ratio ; Wavelet Analysis

Heart sound is a physiological parameter with individual characteristics generated by heart beat. To do the individual classification and recognition, in this paper, we present our study of using wavelet transform in the signal denoising, with the Mel-Frequency cepstrum coefficients (MFCC) as the feature parameters, and propose a research of reducing the dimensionality through principal components analysis (PCA). We have done the preliminary study to test the feasibility of biometric identification method using heart sound. The results showed that under the selected experimental conditions, the system could reach a 90% recognition rate. This study can provide a reference for further research.
Algorithms ; Artificial Intelligence ; Heart Sounds ; physiology ; Humans ; Individuality ; Pattern Recognition, Physiological ; physiology ; Principal Component Analysis ; Signal Processing, Computer-Assisted ; Wavelet Analysis

In this paper, a new method based on the nonlinear chaos theory was proposed to study the arrhythmia with the combination of the correlation dimension and largest Lyapunov exponent, through computing and analyzing these two parameters of 30 cases normal heart sound and 30 cases with arrhythmia. The results showed that the two parameters of the heart sounds with arrhythmia were higher than those with the normal, and there was significant difference between these two kinds of heart sounds. That is probably due to the irregularity of the arrhythmia which causes the decrease of predictability, and it's more complex than the normal heart sound. Therefore, the correlation dimension and the largest Lyapunov exponent can be used to analyze the arrhythmia and for its feature extraction.
Arrhythmias, Cardiac ; diagnosis ; physiopathology ; Heart Sounds ; physiology ; Humans ; Logistic Models ; Nonlinear Dynamics ; Phonocardiography ; Signal Processing, Computer-Assisted

Auscultation, the act of listening for heart sounds to aid in the diagnosis of various heart diseases, is a widely used efficient technique by cardiologists. Since the mechanical prosthetic heart valves are widely used today, it is important to develop a simple and efficient method to detect abnormal mechanical valves. The study on five different mechanical valves showed that only the case of perivalvular leakage could be detected by spectral estimation. Though it is possible to classify different mechanical valves by using time-frequency components of the signal directly, the recognition rate is merely 84%. However, with the improved local discriminant bases (LDB) algorithm to extract features from heart sounds, the recognition rate is 97.3%. Experimental results demonstrated that the improved LDB algorithm could improve classification rate and reduce computational complexity in comparison with original LDB algorithm.
Algorithms ; Heart Sounds ; physiology ; Heart Valve Diseases ; physiopathology ; surgery ; Heart Valve Prosthesis ; Heart Valves ; physiopathology ; Humans ; Pattern Recognition, Automated ; Phonocardiography ; Signal Processing, Computer-Assisted ; Spectrum Analysis ; methods

Exercise Test ; Female ; Heart Sounds ; physiology ; Humans ; Male ; Phonocardiography ; methods ; Pregnancy ; Stress, Physiological ; physiology ; Young Adult

This study was conducted on the basis of informed consent of subjects and was approved by the ethical review committee concerned. 527 pregnant women voluntarily participated in this project for the investigation of cardiac reserve mobilizing condition. Using the digital technique of heart sound signal processing, we measured the heart rate (HR), the ratio of the first heart sound to the second heart sound (S1/S2)and the ratio of diastolic to systolic duration (D/S) during pregnancy. There was significant difference of HR and S1/S2 between the group of non-pregnant (G1), the group of 28-36 pregnant weeks (G2), and the group of 37-42 pregnant weeks (G3) (HR, S1/S2: G1 vs G2, G1 vs G3: P < 0.01). HR of the pregnant women increased with the increase of pregnant weeks. D/S decreased with the increase of pregnant weeks. There was significant difference of D/S between G1, G2, and G3 (G1 vs. G2: P < 0.01; G1 vs. G3: P < 0.01). There was also significant difference of D/S between G2 and G3 (G2 vs. G3: P < 0.05). Everybody in the non-pregnant women group was found to have D/S > or = 1.30; 64.33% of pregnant women were found to have 1.00 < D/S < 1.30, whereas 3.05% of pregnant women were found to have D/S < 1.00. These data revealed that the heart burden of the pregnant woman increased with the increase of pregnant weeks and the mobilization of cardiac reserve.
Adult ; Female ; Heart ; physiology ; Heart Rate ; physiology ; Heart Sounds ; Humans ; Phonocardiography ; methods ; Pregnancy ; physiology ; Pregnancy Trimester, Third ; physiology ; Signal Processing, Computer-Assisted ; Young Adult

BACKGROUNDAlthough acute congestive heart failure (CHF) patients typically present with abnormal auscultatory findings on lung examination, lung sounds are not normally subjected to rigorous analysis. The goals of this study were to use a computerized analytic acoustic tool to evaluate lung sound patterns in CHF patients during acute exacerbation and after clinical improvement and to compare CHF profiles with those of normal individuals.

METHODSLung sounds throughout the respiratory cycle was captured using a computerized acoustic-based imaging technique. Thirty-two consecutive CHF patients were imaged at the time of presentation to the emergency department and after clinical improvement. Digital images were created, geographical area of the images and lung sound patterns were quantitatively analyzed.

RESULTSThe geographical areas of the vibration energy image of acute CHF patients without and with radiographically evident pulmonary edema were (67.9 +/- 4.7) and (60.3 +/- 3.5) kilo-pixels, respectively (P < 0.05). In CHF patients without and with radiographically evident pulmonary edema (REPE), after clinical improvement the geographical area of vibration energy image of lung sound increased to (74.5 +/- 4.4) and (73.9 +/- 3.9) kilo-pixels (P < 0.05), respectively. Vibration energy decreased in CHF patients with REPE following clinical improvement by an average of (85 +/- 19)% (P < 0.01).

CONCLUSIONSWith clinical improvement of acute CHF exacerbations, there was more homogenous distribution of lung vibration energy, as demonstrated by the increased geographical area of the vibration energy image. Lung sound analysis may be useful to track in acute CHF exacerbations.

Adult ; Aged ; Diagnosis, Computer-Assisted ; Female ; Heart Failure ; complications ; Humans ; Male ; Middle Aged ; Pulmonary Edema ; etiology ; pathology ; Respiratory Sounds ; physiology

In order to discriminate normal and abnormal heart sounds accurately and effectively, a new method is proposed to analyze heart sounds, namely heart sound characteristic waveform (HSCW) method. Digital stethoscope is used to collect heart sound signals. The recorded data are transmitted to a computer by USB interface for analysis based on HSCW, which is extracted from an analytical model of single degree-of-freedom (SDOF). Furthermore, a case study on the normal and abnormal cardiac sounds is demonstrated to validate the usefulness and efficiency of the proposed HSCW method. Besides, in order to test the accuracy of discriminating normal and abnormal heart sounds, 40 normal and 20 abnormal heart sounds are collected and analyzed, the accuracy performances are achieved by 92.5% and 95.0%, respectively.
Algorithms ; Heart Auscultation ; instrumentation ; methods ; Heart Sounds ; physiology ; Humans ; Phonocardiography ; methods ; Signal Processing, Computer-Assisted

In this paper is presented the design of detection circuit and single chip acquisition circuit for heart sound, blood pressure and pulse wave. The multi-physiological signals from single chip are synthesized and processed by master computer. The master computer acquires, analyzes and displays the multi-signals by Software LabVIEW. The three physiological signals are fused to calculate the artery stiffness index by time relation of multi-physiological signals. The experiment shows that the circuit is reliable and effective so that it can be used for human vessel function evaluation.
Arteriosclerosis ; diagnosis ; Blood Pressure ; physiology ; Equipment Design ; methods ; Heart Sounds ; physiology ; Humans ; Monitoring, Physiologic ; instrumentation ; methods ; Pulse ; Signal Processing, Computer-Assisted