Claim Missing Document
Check
Articles

CONVOLUTIONAL NEURAL NETWORK FOR ANEMIA DETECTION BASED ON CONJUNCTIVA PALPEBRAL IMAGES Rita Magdalena; Sofia Saidah; Ibnu Da’wan Salim Ubaidah; Yunendah Nur Fuadah; Nabila Herman; Nur Ibrahim
Jurnal Teknik Informatika (Jutif) Vol. 3 No. 2 (2022): JUTIF Volume 3, Number 2, April 2022
Publisher : Informatika, Universitas Jenderal Soedirman

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.20884/1.jutif.2022.3.2.197

Abstract

Anemia is a condition in which the level of hemoglobin in a person's blood is below normal. Hemoglobin concentration is one of the parameters commonly used to determine a person's physical condition. Anemia can attack anyone, especially pregnant women. Currently, many non-invasive anemia detection methods have been developed. One of non-invasive methods in detecting anemia can be seen through its physiological characteristics, namely palpebral conjunctiva images. In this study, conjunctival image-based anemia detection was carried out using one of the deep learning methods, namely Convolutional Neural Netwok (CNN). This CNN method is used with the aim of obtaining more specific characteristics in distinguishing normal and anemic conditions based on the image of the palpebral conjunctiva. The Convolutional Neural Network proposed model in this study consists of five hidden layers, each of which uses a filter size of 3x3, 5x5, 7x7, 9x9, and 11x11 and output channels 16, 32, 64, 128 respectively. Fully connected layer and sigmoid activation function are used to classify normal and anemic conditions. The study was conducted using 2000 images of the palpebral conjunctiva which contained anemia and normal conditions. Furthermore, the dataset is divided into 1,440 images for training, 160 images for validation and 400 images for model testing. The study obtained the best accuracy of 94%, with the average value of precision, recall and f-1 score respectively 0.935; 0.94; 0.935. The results of the study indicate that the system is able to classify normal and anemic conditions with minimal errors. Furthermore, the system that has been designed can be implemented in an Android-based application so that the detection of anemia based on this palpebral conjunctival image can be carried out in real-tim.
Modifikasi Convolutional Neural Network Arsitektur GoogLeNet dengan Dull Razor Filtering untuk Klasifikasi Kanker Kulit Sofia Saidah; I Putu Yowan Nugraha Suparta; Efri Suhartono
Jurnal Nasional Teknik Elektro dan Teknologi Informasi Vol 11 No 2: Mei 2022
Publisher : Departemen Teknik Elektro dan Teknologi Informasi, Fakultas Teknik, Universitas Gadjah Mada

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (1265.14 KB) | DOI: 10.22146/jnteti.v11i2.2739

Abstract

Skin is the widest external organ covering the human body. Due to a high intensity exposure to the environment, the skin can experience various health problems, one of which is skin cancer. Early detection is needed so that further medication for patients can be done immediately. In this regard, the use of artificial intelligence (AI)-based solutions in detecting skin cancer images can be used to detect skin cancer potentials. In this study, the classification of benign and malignant skin cancer types was carried out by utilizing GoogLeNet Convolutional Neural Network (CNN) method. The GoogLeNet architecture has the advantage of having an inception module, allowing the convolution and pooling processes to run in parallel terms that can reduce computing time and speed up the classification process without lowering the system accuracy. This study consisted of several stages, starting from the data acquisition of 600 skin cancer images from Kaggle.com to the uniformity of the input size that allows the system to work faster. There was also a utilization of dull razor filtering to reduce input image noise due to hair growing along the epidermis. After the preprocessing process was complete, GoogLeNet architecture processed the image input before categorizing the input into benign or malignant skin cancer. The system’s performance was then evaluated using performance parameters such as accuracy, precision, recall, and F-1 score, and it was compared to other methods. The system managed to obtain satisfactory results, including the accuracy of 97.73% and the loss of 1.7063. As for precision, recall, and F-1 score parameters, each received an average value of 0.98. The system performance proposed by the authors successfully have better accuracy compared to the previous study with much less use of datasets. The test results show that CNN method is able to detect and classify skin cancer accurately, so it is expected that this method could help medical workers in diagnosing the community.
GLAUCOMA CLASSIFICATION BASED ON FUNDUS IMAGES PROCESSING WITH CONVOLUTIONAL NEURAL NETWORK Yunendah Nur Fu'adah; Sofia Saidah; Nidaan Khofiya; Rita Magdalena; Ibnu Da'wan Salim
Jurnal Teknik Informatika (Jutif) Vol. 3 No. 3 (2022): JUTIF Volume 3, Number 3, June 2022
Publisher : Informatika, Universitas Jenderal Soedirman

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.20884/1.jutif.2022.3.3.276

Abstract

Glaucoma is an eye disease that causes damage to the optic nerve due to increased pressure in the eyeball. Delay in diagnosis and treatment of optic nerve damage due to glaucoma can lead to permanent blindness. Thus, several studies have developed a glaucoma early detection system based on digital image processing and machine learning. This study carried out glaucoma classification based on fundus image processing using Convolutional Neural Network (CNN). The CNN architecture proposed in this study consists of three convolutional layers with output channels 8, 16, 32 sequentially and a filter size of 5×5 at each layer, followed by a pooling layer and a dropout layer at the feature extraction stage. Furthermore, a fully connected layer and softmax activation function was implemented at the classification stage to classify fundus images into normal conditions, early glaucoma, moderate glaucoma, deep glaucoma, and ocular hypertension (OHT). The total amount of fundus image data used in this study consisted of 2000 fundus images divided into 1280 training data, 320 validation data, and 400 test data. 5-fold cross-validation is implemented in the training phase to select the best model. At the testing stage, the best accuracy generated by 99%, with the precision value, recall, f-1 scores and the AUC score are close to 1. According to the system performance results obtained, the proposed model can be used as a tool for medical personnel in classifying glaucoma conditions to provide appropriate medical treatment and reduce the risk of permanent blindness due to glaucoma.
Analisis Perbandingan K-Nearest Neighbor dan Support Vector Machine pada Klasifikasi Jenis Sapi dengan Metode Gray Level Coocurrence Matrix Salsabil Farah Aqilah Wijaya; Koredianto Koredianto; Sofia Saidah
Jurnal Ilmu Komputer dan Informatika Vol 2 No 2 (2022): JIKI - Desember 2022
Publisher : CV Firmos

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.54082/jiki.27

Abstract

Sapi merupakan hewan ternak yang banyak dibudidayakan di Indonesia mulai dari daging, susu, kotoran, kulit, hingga membantu bercocok tanam. Menurut Badan Pusat Statistik pada tahun 2020 terdapat 17.466.792 ekor populasi sapi potong yang ada di Indonesia. Dari 17.466.792 ekor populasi sapi potong yang ada di Indonesia terdapat 896.200 ekor populasi sapi yang ada di Sumatera Utara yang merupakan provinsi 6 teratas yang memiliki populasi sapi potong terbanyak. Tetapi permasalahannya masih banyak peternak yang tidak mengetahui jenis dari sapi yang dimiliki sehingga perawatan yang salah pada sapi yang dimiliki tentu akan berpengaruh terhadap kualitas sapi yang dihasilkan. Dalam penelitian merancang sistem klasifikasi jenis sapi dengan metode Gray Level Cooccurrence Matrix (GLCM) menggunakan klasifikasi K-Nearest Neighbor dan Support Vector Machine (SVM) menggunakan 2 jenis klasifikasi yaitu klasifikasi K-NN dan klasifikasi SVM. Dalam pengujian ini didapatkan akurasi sebesar 100% pada klasifikasi K-NN dengan waktu komputasi sebesar 0.967 s dengan menggunakan jenis distance mahalonobis dengan nilai k =1 dan pada klasifikasi SVM didapatkan tingkat akurasi 80% dengan waktu komputasi sebesar 1.570 s dengan menggunakan jenis kernel polynomial dengan kelas SVM OAO. Dari hasil pengujian yang didapatkan sistem klasifikasi jenis sapi lebih mendapatkan nilai akurasi terbaik pada klasifikasi K-NN dengan nilai K=1 dan jenis distance mahalanobis.
DESIGN AND IMPLEMENTATION OF LEARNING TOOLS TO READ THE BRAILLE LETTERS BASED ON VOICE PROCESSING AND ARDUINO USING MEL FREQUENCY CEPSTRAL COEFFICIENT AND K-NEAREST NEIGHBOR METHOD Raditiana Patmasari; Sofia Saidah; A F Akbar; Rita Magdalena
JMECS (Journal of Measurements, Electronics, Communications, and Systems) Vol 6 No 1 (2020): JMECS
Publisher : Universitas Telkom

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.25124/jmecs.v6i1.2019

Abstract

Ability to read Braille is critical skill for blind students. Without the skill, blind students would encounter difficulties in their learning activities because most learning materials are written using the Braille system. The currently applied Braille learning system uses printed paper that is time consuming and pricey. This research attempts to develop a tool for helping the blinds to learn how to read braille letters. The tool processes inputs in the form of speech signal into a text by applying Mel Frequency Cepstral Coefficient (MFCC) as a feature extraction method and K- Nearest Neighbor (KNN) as a classifier method. The text will subsequently be transformed into Braille pattern by using Arduino UNO. The test results discover the combination of Mel Frequency Cepstral Coefficient and K-Nearest Neighbor method are able to recognize the speech signal of different alphabets with 87,3% accuracy. Furthermore, the computing time for alphabet recognitions decreases 85 % when the device is applied This finding helps the blind students to recognize the alphabets easily and faster.
Dual Steganography in Digital Images with Spread Spectrum Insertion Method Sofia Saidah
JMECS (Journal of Measurements, Electronics, Communications, and Systems) Vol 5 No 1 (2019): JMECS
Publisher : Universitas Telkom

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.25124/jmecs.v5i1.2073

Abstract

This study aims to prove whether embedding a stego-image within another cover can be performed to deceive hackers or unauthorized people. In steganography, both concealing the fact that secret message is sent and its content are concerned. Dual steganography means to make unauthorized people think that the first cover is the real message. The first step is to embed the secret text message into the first cover image using Spread Spectrum (SS) method. After that, the resulting stego-image was transformed using the Discrete Wavelet Transform (DWT) method followed by the insertion process using the Singular Value Decomposition (SVD) method. The result of this study shows that the system can perform dual steganography with good imperceptibility. Parameters measured in the average of 35 dB of PSNR and 30 dB of SNR in the first embedding process; meanwhile, for the second process the system performed in the average of 41 dB of PSNR and 38 dB of SNR. Also, in the extraction process, BER measured close to 0. Although some basic attack scenarios such as Gaussian noise, Salt and Pepper noise, Low Pass Filter (LPF) and High Pass Filter (HPF) were performed in this research, more advanced attack scenarios can be discussed in future research; for instance, compression and geometric transform.
Individual Identification Through Voice Using Mel-Frequency Cepstrum Coefficient (MFCC) and Hidden Markov Models (HMM) Method Dea Sifana Ramadhina; Rita Magdalena; Sofia Saidah
JMECS (Journal of Measurements, Electronics, Communications, and Systems) Vol 7 No 1 (2020): JMECS
Publisher : Universitas Telkom

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.25124/jmecs.v7i1.3553

Abstract

Voice is one of the parameters in the identification process of a person. Through the voice, information will be obtained such as gender, age, and even the identity of the speaker. Speaker recognition is a method to narrow down crimes and frauds committed by voice. So that it will minimize the occurrence of faking one's identity. The Method of Mel Frequency Cepstrum Coefficient (MFCC) can be used in the speech recognition system. The process of feature extraction of speech signal using MFCC will produce acoustic speech signal. The classification, Hidden Markov Models (HMM) is used to match unidentified speaker’s voice with the voices in database. In this research, the system is used to verify the speaker, namely 15 text dependent in Indonesian. On testing the speaker with the same as database, the highest accuracy is 99,16%.
INDIVIDUAL IDENTIFICATION BY IRIS USING HISTOGRAM OF ORIENTED GRADIENT (HOG) AND BACKPROPAGATION NEURAL NETWORK Widya Alisya Kusuma Ningrum; Iwan Iwut Tritoasmoro; Sofia Saidah
Jurnal Teknik Informatika (Jutif) Vol. 4 No. 2 (2023): JUTIF Volume 4, Number 2, April 2023
Publisher : Informatika, Universitas Jenderal Soedirman

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.52436/1.jutif.2023.4.2.558

Abstract

The eye’s iris biometrics is a type of biometric for individual identification that is more stable than other types of biometrics because a person's iris eye’s has a delicate fiber pattern and unique characteristics. Especially with the rapid development of the times, the need for identity recognition systems is also increasing. Introducing individuals in traditional ways is still less effective than biometric systems because, compared to conventional methods, biometric systems are safer and are not easily stolen, imitated, or accessed by any unauthorized person. In this research has been carried out by designing a simulation system for individual identification through iris eyes images using the Histogram of Oriented Gradien (HOG) method for image extraction. They were continued with classification using Artificial Neural Network (ANN) Backpropagation. The dataset used is primary data taken directly through smartphone cameras from 30 individuals.Based on the test results and analysis of the Histogram of Oriented Gradien method using an image size of 128×128 pixels, parameters of Cell Size 16×16 cells, Bins Numbers 12, Size Block 2×2 cells, L2-Hys normalization scheme, and JST backpropagation classification with Random state value 1, Learning Rates 0.001, Epoch 200, Hidden Layer 100 with the system's sigmoid activation function can produce a performance system with the most significant performance accuracy of 91.93% , using 1500 training data and 1500 iris eyes image test data.
TONE DETECTION ON TERANIKA MUSICAL INSTRUMENT USING DISCRETE WAVELET TRANSFORM AND DECISION TREE CLASSIFICATION Fadia Qothrunnada; Sofia Saidah; Bambang Hidayat; Tasya Busrizal Putri; Darwindra
Jurnal Teknik Informatika (Jutif) Vol. 4 No. 2 (2023): JUTIF Volume 4, Number 2, April 2023
Publisher : Informatika, Universitas Jenderal Soedirman

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.52436/1.jutif.2023.4.2.570

Abstract

Musical instruments are one of the cultures that must be preserved. Teranika is one of the traditional musical instruments from the Majalengka area, which is made of clay. Currently, the manufacture of conventional musical instruments is still done manually, so there are still differences in the tone produced. Meanwhile, the quality of a musical instrument is determined by the accuracy of the technique. Therefore, we need a system that can accurately detect the method's accuracy. The author designed a tone detection system for Teranika musical instruments to help artisans carry out quality control. This system will detect whether or not the musical instrument is successfully matched with the right tone and agent. The technique contained in this musical instrument is Do Re Mi Fa So La Si Do high. To overcome these problems, the author makes this tone detection system using the Discrete Wavelet Transform method and the Decision Tree classification. The working principle of this system is that the recorded sound of musical instruments will be transmitted to this system. Then the sound will be processed as input and matched with the essential voice in the database. The output of this system produces samples according to the sampling frequency used. The test results show the best results at decomposition level 6, a thresholding value of 0.05, and a Fine Tree classification type with an accuracy of 87.5%
Audio Watermarking Dengan Menggunakan Metode Fast Fourier Transform (fft) Dan Singular Value Decomposition (svd) Robinzon Pakpahan; Ratri Dwi Atmaja; Sofia Saidah
eProceedings of Engineering Vol 5, No 2 (2018): Agustus 2018
Publisher : eProceedings of Engineering

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

ABSTRAK Perkembangan teknologi yang semakin maju membuat memudahkan orang dalam melakukan duplikasi dan menyebarkan konten audio digital tanpa sepengetahuan pemilik, khususnya pada dunia digital. Hal ini membuat dirancang Watermarking yang digunakan untuk melindungi hak cipta pemilik khususnya pada data audio. Pada penelitian ini dirancang dan direalisasikan Watermarking untuk audio digital untuk menjaga hak cipta pemilik audio dengan melakukan penyisipan citra informasi kedalam suatu file audio dan ekstraksi audio ter-Watermark dengan menggunakan metode FFT (Fast Fourier Transform) dan SVD (Singular Value Decomposition) Metode FFT (Fast Fourier Transform) merupakan pembawa suatu citra dari ruang spasial ke ruang frekuensi dan metode SVD (Singular Value Decomposition) merupakan salah satu alat matematis yang di gunakan untuk merepresentasikan sebuah matrik dan mampu melakukan berbagai analisis dan komputasi. Berdasarkan audio Watermarking, audio Watermarking dengan metode Fast Fourier Transform dan Singular Value Decomposition dapat menyisipkan citra dan menghasilkan performansi audio terWatermark dengan nilai terbaik BER 0%, SSIM 1, dan SNR tertinggi 69db. Kata kunci: Audio Watermarking, Fast Fourier Transform, Singular Decomposition Value ABSTRACT The development of increasingly advanced technology makes it easier for people to duplicate and distribute digital audio content without the owner's knowledge, especially in the digital world. This makes designed watermarking used to protect copyright owners in particular in the audio data. In this study, designed and realized watermarking for digital audio to keep the copyright owner of audio by inserting image information into an audio file and extract audio ter-Watermark by using FFT (Fast Fourier Transform) and SVD (Singular Value Decomposition) The FFT (Fast Fourier Transform) method is the carrier of an image from spatial space to the frequency space and the SVD method (Singular Value Decomposition) is one of the mathematical tools used to represent a matrix and is capable of performing various analyzes and computations. Based on the Watermarking audio, Watermarking audio with Fast Fourier Transform and Singular Value Decomposition methods can insert images and produce the best Watermark audio performance with the best BER 0%, SSIM 1, and 69db highest SNR values. Keywords: Audio Watermarking, Fast Fourier Transform, Singular Decomposition Value
Co-Authors A F Akbar Abel Bima Wiratama Aditya S.B, I Dewa Agung Akbar Trisnamulya Putra Al Brando Ardes Harjoko Aliefiya Rachman Alif Fajri Ryamizard Alrizqi, Naufal Dwi Andre Megantoro, Andre Megantoro Angga Prihantoro Arfat, Ikrar Khaera ARIS HARTAMAN Azzahra, Fatima Bainuri, Aulia Novria Bambang Hidayat Bambang Hidayat Bambang Hidayat Bongso, Dery Febryanto Darwindra Darwindra, Darwindra Dea Sifana Ramadhina Denny Darlis Desi Dwi Prihatin Dyah Ajeng Pramudhita Effendi , Doni Oktavian Ibnu Efri Suhartono Enrico Wiratama Purwanto Fadia Qothrunnada Fardiyanti, Defitriana Fathurrahman, Muhammad Hanif Fatima Azzahra Fellia Rizki Kusumowardani Fiera Meiristika Utami Firdaus, Muhammad Ilham Zuhruf Fitria, Ismaulida Nur Gaol, Satya Wira Fernanda Lumban Gelar Budiman Hilman Fauzi, Hilman Hurianti Vidya I Putu Yowan Nugraha Suparta Ibnu Da'wan Salim Ibnu Da’wan Salim Ubaidah Ibnu Da’wan Salim Ubaidah Ikhwanda, Alfan Ikrar Khaera Arfat Inung Wijayanto Iqbal Kurniawan Perdana Israndy Yainahu Iwan Iwut Tritoasmoro Jangkung Raharjo Kintan Veriana Koredianto Koredianto Koredianto Usman Mas, Muhammad Sabri Masykur, Muhammad Fadhel Affandi Muhamad Rokhmat Isnaini Muhammad Bayu Adinegara Muhammad Ilham Muhammad Ilham Muhammad, Zalfa Alif Nabila Herman Nidaan Khofiya Nor Kumalasari Caecar Nor Kumalasari Caecar Pratiwi Nur Alyyu Nur Ibrahim Perdana, Iqbal Kurniawan Pramudhita, Dyah Ajeng PRATIWI, NOR KUMALASARI CAESAR Prayudi, Yoshi Prihantoro, Angga Putra, Akbar Trisnamulya Putri , Yusnita Putri, Tasya Busrizal Putri, Yusnita Qothrunnada, Fadia R. Yunendah Nur Fu’adah Rachmat Hidayat Ashary Raditiana Patmasari Ratna Sari Ratri Dwi Atmaja Reza Ahmad Nurfauzan Richard Bina Jadi Simanjuntak Rita Magdalena Rita Magladena Rita Purnamasari Robinzon Pakpahan Salsabil Farah Aqilah Wijaya Salsabila, Afap Sangkala, Muh Aslam Mahdi Sevierda Raniprima Subiakto, Septiaini Dela Susilo, Mochammad Hilmi Suwandhi, Adhisty Putrina Syamsul Rizal Syamsul Rizal Tahta Restu Adiguna Tasya Busrizal Putri Tita Haryanti Tsabita Al Asshifa Hadi Kusuma Vidya, Hurianti Wahid, Gloria Shekinah Florensia Wibisono Sabdo Utomo WIDIANTO, MOCHAMMAD HALDI Widya Alisya Kusuma Ningrum Yunendah Fu’adah Zakiah Zakiah