Klasifikasi Suara Jantung Menggunakan Convolutional Neural Network 1 Dimensi (Cnn 1d) Dengan Pitch Shifting Dan Transformasi FFT Pada Dataset Terbatas

Q: What is the clinical evidence from this study?

Population: Heart sound classification (normal vs murmur). Intervention: 1D Convolutional Neural Network (CNN) with pitch shifting and FFT. Primary outcome: Model classification performance (accuracy, precision, recall, F1 score).

Q: What does this research mean for the field?

A 1D convolutional neural network utilizing Butterworth filtering, pitch-shifting augmentation, and FFT feature extraction classifies normal and murmur heart sounds from phonocardiogram signals with an average accuracy of 76.15%. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Edinar Valiant Hawali; Reza Bagus Acintyasakti; Dwi Definta Oktavia Siswoyo

doi:10.37303/sistem.v22i1.344

April 7, 2026SISTEM Jurnal Ilmu Ilmu TeknikOpen Access

Klasifikasi Suara Jantung Menggunakan Convolutional Neural Network 1 Dimensi (Cnn 1d) Dengan Pitch Shifting Dan Transformasi FFT Pada Dataset Terbatas

View Full Paper

Ask AI

Bookmark

Share

Key result

A 1D Convolutional Neural Network with pitch shifting and FFT achieved an average accuracy of 0.7615, precision of 0.8533, and F1 score of 0.7517 for classifying normal and murmur heart sounds.

Why the study?

Early detection of heart abnormalities is important to prevent serious complications, but effective classification methods for heart sounds using limited datasets are needed.

Does a 1D CNN model combined with pitch shifting and FFT accurately classify normal heart sounds versus murmurs using a limited PCG dataset?

Key Points

Classify normal heart sounds and cardiac murmurs from limited phonocardiogram datasets using a one-dimensional convolutional neural network.
Filtered phonocardiogram signals using a fourth-order Butterworth bandpass filter (30–150 Hz) and extracted frequency features via Fast Fourier Transform.
Augmented limited sound data using pitch shifting (+2 semitones) and evaluated a 1D CNN architecture through 5-fold cross-validation.
The 1D CNN model achieved an average accuracy of 0.7615 and an F1 score of 0.7517 across cross-validation folds.
The classifier yielded a precision of 0.8533 and a recall of 0.6912 when identifying heart murmurs compared to normal sounds.

Structured PICO

Does a 1D CNN model combined with pitch shifting and FFT accurately classify normal heart sounds versus murmurs using a limited PCG dataset?

P

Population

Phonocardiogram (PCG) signals from a limited dataset used for classifying normal heart sounds and murmurs.

I

Intervention

1-Dimensional Convolutional Neural Network (1D CNN) classification model with preprocessing (4th order Butterworth band-pass filter 30-150 Hz), +2 semitone pitch shifting augmentation, and Fast Fourier Transform (FFT)

O

Outcome

Classification performance (accuracy, precision, recall, F1 score) between normal heart sounds and murmurs

Limitations

Limited dataset size
Limited dataset

Implication

A 1-dimensional CNN utilizing signal preprocessing, pitch shifting augmentation, and FFT can effectively classify normal heart sounds and murmurs from PCG signals, achieving 76.15% average accuracy.

Abstract

Pendeteksian dini terhadap kelainan jantung sangat penting untuk mencegah risiko komplikasi yang lebih serius. Penelitian ini bertujuan untuk mengklasifikasikan suara jantung normal dan murmur menggunakan metode Convolutional Neural Network (CNN) satu dimensi. Data yang digunakan merupakan sinyal phonocardiogram (PCG) yang telah melalui tahapan preprocessing berupa band pass filter Butterworth orde 4 (30 – 150 Hz), augmentasi pitch shifting (+2 semitone), dan transformasi Fast Fourier Transform (FFT) untuk ekstraksi fitur dalam domain frekuensi. Model CNN dilatih menggunakan skema validasi silang 5 fold untuk memastikan performa yang stabil di berbagai subset data. Analisis bentuk gelombang, spektrum frekuensi, dan spektrogram menunjukkan adanya perbedaan signifikan antara suara jantung normal dan murmur, serta dampak dari proses augmentasi terhadap distribusi frekuensi. Hasil evaluasi menunjukkan bahwa model mampu mencapai rata-rata akurasi sebesar 0,7615, dengan nilai precision sebesar 0,8533, recall sebesar 0,6912, dan F1 score sebesar 0,7517. Penelitian ini menunjukkan bahwa kombinasi preprocessing sinyal, augmentasi data, dan CNN satu dimensi dapat digunakan secara efektif untuk klasifikasi suara jantung. Hasil ini berpotensi untuk dikembangkan lebih lanjut dalam sistem deteksi dini gangguan jantung berbasis PCG.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Hawali et al. (2026) studied Heart sound classification (normal vs murmur). 1D Convolutional Neural Network (CNN) with pitch shifting and FFT was evaluated on Model classification performance (accuracy, precision, recall, F1 score). A 1D Convolutional Neural Network with pitch shifting and FFT achieved an average accuracy of 0.7615, precision of 0.8533, and F1 score of 0.7517 for classifying normal and murmur heart sounds.

synapsesocial.com/papers/6ac71bf0ce9b52c5cbf931aa https://doi.org/10.37303/sistem.v22i1.344

View Full Paper

Ask AI

Bookmark

Share

Klasifikasi Suara Jantung Menggunakan Convolutional Neural Network 1 Dimensi (Cnn 1d) Dengan Pitch Shifting Dan Transformasi FFT Pada Dataset Terbatas

Key Points

Structured PICO

Limitations

Implication

Abstract

Discussion

Cite This Study

Klasifikasi Suara Jantung Menggunakan Convolutional Neural Network 1 Dimensi (Cnn 1d) Dengan Pitch Shifting Dan Transformasi FFT Pada Dataset Terbatas

Key Points

Structured PICO

Limitations

Implication

Abstract

Discussion

Cite This Study

Keep reading

More related papers

Keep reading

More related papers