Pratama, Dionisius
Unknown Affiliation

Published : 2 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 2 Documents
Search

Penerapan Convolutional Neural Network untuk Melakukan Estimasi Pitch pada Rekaman Suara Penyanyi Pratama, Dionisius; Heryanto, Hery; Kurniawan, Hans Christian
Jurnal Telematika Vol. 16 No. 2 (2021)
Publisher : Yayasan Petra Harapan Bangsa

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.61769/telematika.v16i2.396

Abstract

A musical performance is determined by the intonation accuracy, which is the pitch accuracy of a musician or musical instrument, whether a tone is played 'in tune' or not. Therefore, to determine the intonation quality of a musical performance, it is necessary to estimate the pitch. In this research, a one-dimensional Convolutional Neural Network (CNN) is used to estimate the pitch from singing voice recording. After pitch estimation, Dynamic Time Warping (DTW) method is used to calculate the similarity (measured in distance) of pitch estimation results with the recording template from the dataset to determine intonation accuracy. This research uses several preprocessing methods, such as quantization pitch label, spectrogram generation, scaling, and spectrogram recoloring. The CNN method for performing pitch estimation is tested using five songs from the MIR-QBSH dataset. CNN testing is done by applying four architectural designs by combining epoch values, learning rate, number of filters in each convolutional layer, and number of convolutions to find the best combination that produces the highest accuracy. Based on the test results, the model built can produce the highest average accuracy of 97.425% with a difference between the average accuracy and the average validation accuracy of 14.383%. The optimal threshold value for distance is in the range of 1000-1500.  Pembawaan karya musik yang baik ditentukan dari ketepatan intonasi yang merupakan akurasi pitch dari sebuah nada yang dikeluarkan oleh seorang musisi atau instrumen musik, diproduksi dengan tepat atau tidak. Maka dari itu, untuk menentukan kualitas intonasi penampilan suatu karya musik, estimasi pitch perlu dilakukan. Pada penelitian ini, sebuah Convolutional Neural Network (CNN) satu dimensi digunakan untuk melakukan estimasi pitch dari rekaman suara nyanyian. Setelah estimasi pitch dilakukan, maka digunakan metode Dynamic Time Warping (DTW) untuk melakukan pengujian kemiripan (dalam distance) hasil estimasi pitch dengan template rekaman dari dataset. Pengujian tersebut dilakukan untuk menentukan ketepatan intonasi. Beberapa metode preprocessing yang dilakukan adalah pembulatan pitch label, pembuatan spektogram, scaling, dan pewarnaan ulang spektogram. Metode CNN untuk melakukan estimasi pitch diuji dengan menggunakan lima lagu dari dataset MIR-QBSH. Pengujian CNN dilakukan dengan menerapkan empat rancangan arsitektur dengan mengombinasikan nilai epoch, learning rate, jumlah filter pada setiap convolutional layer, dan jumlah konvolusi untuk mencari kombinasi terbaik yang menghasilkan akurasi tertinggi. Berdasarkan hasil pengujian, model yang dibangun dapat menghasilkan rata-rata akurasi tertinggi sebesar 97,425% dengan selisih antara rata-rata akurasi dan rata-rata akurasi validasi sebesar 14,383%. Nilai threshold yang optimal untuk distance berada pada rentang 1000-1500.
Rancangan Dasar Sistem Aplikasi Pemantau Lalu Lintas dan Penghitung Kendaraan Berbasis Komputasi Tepi Heryanto, Hery; Hutagalung, Maclaurin; Gamaliel, Yoyok Yusman; Angela, Dina; Pratama, Dionisius; Martina, Inge; Nugroho, Tunggul Arief
JURNAL INFOTEL Vol 16 No 2 (2024): May 2024
Publisher : LPPM INSTITUT TEKNOLOGI TELKOM PURWOKERTO

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.20895/infotel.v16i2.1105

Abstract

One of the main issues in Indonesia is congestion. The number of vehicles continues to increase and is less balanced by the development of transportation infrastructure, especially landlines, causing more complex problems. The Indonesian government needs an intelligent application system that can provide knowledge to unravel congestion. The problem is how to perform edge computing to reduce latency so that the highway monitoring application system runs in real time. This research proposes a basic design for a vehicle monitoring application system that can accurately recognize vehicles, count the number of vehicles, and propose an edge computation that brings computation directly to the data source. The dataset is a video of traffic in Bandung, Jakarta, and several other major cities. The images in the dataset consist of 4,890 training images, 467 validation images, and 231 testing images. In the proposed model, the YOLOv5 and YOLOv7 architectures accurately detect and count vehicles. The test results show a mAP value of 99.1% with an IoU threshold of 50%. Other results include a precision value of 96.2% and a recall of 97.7%. The proposed model can accurately monitor vehicles and reduce latency with an edge computing approach.