ANALYSIS OF HEARTBEAT ANOMALIES FROM DIGITAL STETHOSCOPE AUDIO
Accurate manual diagnosis of heart diseases by physician from audio stethoscope signals requires special training and depends on the skill of the physicians. With the increasing number of the cardiac problems, the importance of efficient determination of the specific heart condition is required....
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Year Project |
| Language: | English |
| Published: |
2018
|
| Online Access: | http://utpedia.utp.edu.my/19171/1/Hussein%20Baashen_20543_FYP_Final%20Report%20copy.pdf http://utpedia.utp.edu.my/19171/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Accurate manual diagnosis of heart diseases by physician from audio stethoscope
signals requires special training and depends on the skill of the physicians. With the
increasing number of the cardiac problems, the importance of efficient determination
of the specific heart condition is required. There are a lot of ways to classify a heart
diagnosis other than using medical examiner’s hearing ability. This work presents a
research on analyzing the heartbeat anomalies using digital stethoscope audio. In order
to validate the classification of different heartbeat anomalies results, the proposed
framework will be applied on publicly available standard heart sound dataset. The
heart sound can be classified into 4 different heart anomaly, namely normal, extra heart
sounds, artifact and murmur. Firstly, noise removal method based on the Savitzky-
Golay (SG) filter is used to remove the noise disturbance in the signal. And then a total
of 19 features are extracted using the Wavelet Packet Decomposition (WPD) of level
2 and dB 3 decomposition and additional 10 features are using the Short-Time Fourier
Transform (STFT). The features extracted using WPD are energy, entropy, mean,
standard deviation and covariance, interquartile rate, mean absolute deviation,
skewness, kurtosis, median from the decomposed wavelet sub bands and 10 features
spectrogram using STFT method. And the final stage is to get the classification
accuracy using all methods set in the classification learner application in MATLAB to
find the most accurate method with the highest accuracy rate. The best method for
classifying the heartbeat anomalies is the Ensemble Methods, using the Bagged submethod.
The highest classification accuracy percentage for this project is 83.8% for 4
classes classification. |
|---|