Haryati Jaafar
Thumbnail Image
Preferred name
Haryati Jaafar
Official Name
Jaafar, Haryati
Alternative Name
Jaafar, Haryati
Jaafar, H.
Main Affiliation
Scopus Author ID
55357649900
Researcher ID
ILC-1943-2023

Research Output

Filters

7 1
1 1
7
Reset filters

Settings
Now showing 1 - 7 of 7
  • Publication
    Frog sound identification using extended k-nearest neighbor classifier
    ( 2017-09-21)
    Nordiana Mukahar
    ;
    Bakhtiar Affendi Rosdi
    ;
    Dzati Athiar Ramli
    ;
    Haryati Jaafar
    Frog sound identification based on the vocalization becomes important for biological research and environmental monitoring. As a result, different types of feature extractions and classifiers have been employed to evaluate the accuracy of frog sound identification. This paper presents a frog sound identification with Extended k-Nearest Neighbor (EKNN) classifier. The EKNN classifier integrates the nearest neighbors and mutual sharing of neighborhood concepts, with the aims of improving the classification performance. It makes a prediction based on who are the nearest neighbors of the testing sample and who consider the testing sample as their nearest neighbors. In order to evaluate the classification performance in frog sound identification, the EKNN classifier is compared with competing classifier, k -Nearest Neighbor (KNN), Fuzzy k -Nearest Neighbor (FKNN) k - General Nearest Neighbor (KGNN)and Mutual k -Nearest Neighbor (MKNN) on the recorded sounds of 15 frog species obtained in Malaysia forest. The recorded sounds have been segmented using Short Time Energy and Short Time Average Zero Crossing Rate (STE+STAZCR), sinusoidal modeling (SM), manual and the combination of Energy (E) and Zero Crossing Rate (ZCR) (E+ZCR) while the features are extracted by Mel Frequency Cepstrum Coefficient (MFCC). The experimental results have shown that the EKNCN classifier exhibits the best performance in terms of accuracy compared to the competing classifiers, KNN, FKNN, GKNN and MKNN for all cases.
  • Publication
    Deep neural network approach to frog species recognition
    ( 2017-10-10)
    Norsalina Hassan
    ;
    Dzati Athiar Ramli
    ;
    Haryati Jaafar
    Automatic frog species recognition based on acoustic signal has received attention among biologists for environmental studies as it can detect, localize and document the declining population of frog species efficiently compared to the manual survey. In this study, we investigate the possibility of the use of Deep Neural Network (DNN) as a classifier for a frog species recognition system. The Mel-Frequency Cepstral Coefficients (MFCCs) is utilized as features and prior to the feature extraction, we also investigate the capability of automatic segmentation of syllables based on the Sinusoidal Modulation (SM), Energy with Zero Crossing Rate (E+ZCR) and Short-Time Energy with Time Average Zero Crossing Rate (STE+STAZCR). We also evaluate several DNN parameter's setting so as to discover the optimum parameter values for our developed system. 55 different species of frog with 2674 syllables from our in-house database have been tested. Experimental results based on DNN classifier showed that the STE+STAZCR method gives the accuracy of 99.03%, which reveals the viability of DNN as a classifier. In future, further research on DNN parameter optimization will be conducted for system improvement.
  • Publication
    Automatic detection of embolic signal for stroke prevention
    ( 2017-01-01)
    Noor Salwani Ibrahim
    ;
    Ng Yan Duan
    ;
    Dzati Athiar Ramli
    ;
    Haryati Jaafar
    Transcranial Doppler (TCD) ultrasound is an essential tool in clinical diagnosis to determine the occurrence of embolism in stroke patients. However, it requires manual attention and the accuracy will deteriorate due to fatigue factor. Instead of depending on human observer as a gold standard to detect the emboli, this study proposes an automated emboli detection system based on three detection methods i.e. time-domain intensity, frequency-domain intensity and time-frequency intensity hybrid. Experimental studies of 240 samples of six data sets were employed. The performance evaluations of each method are measured in term of accuracy percentage and processing speed while human observation is also done as the golden standard for accuracy comparison. The best result is achieved by the time-frequency intensity hybrid method where 90.74 % of the embolic signals and 100 % of the non-embolic signals were successfully identified. The performance of this method is promising as the accuracy achieved by human observation was 87.45 and 100 % for embolic signals and non-embolic signals, respectively.
  • Publication
    A methodology of nearest neighbor: Design and comparison of biometric image database
    ( 2017-01-06)
    Haryati Jaafar
    ;
    Nordiana Mukahar
    ;
    Dzati Athiar Ramli
    The nearest neighbor (NN) is a non-parametric classifier and has been widely used for pattern classification. Nevertheless, there are some problems encountered that leads to the poor performance of the NN i.e. the samples distribution, weighting issues and computational time for large databases. Hence, various classifiers i.e. k Nearest Neighbor (kNN), k Nearest Centroid Neighborhood (kNCN), Fuzzy k Nearest Neighbor (FkNN), Fuzzy-Based k Nearest Centroid Neighbor (FkNCN) and Improved Fuzzy-Based k Nearest Centroid Neighbor (IFkNCN) were proposed to improve the performance of the NN. This paper presents a review of aforementioned classifiers including the taxonomy, toward the implementation of classifiers in biometric image database. Two databases i.e. finger print and finger vein have been employed and the performance of classifiers were compared in term of processing time and classification accuracy. The results show that the IFkNCN classifier owns the best accuracies to the kNN, kNCN FkNN and FkNCN with 97.66% and 96.74% for fingerprint and finger vein databases, respectively.
  • Publication
    User specific weights based on score distance and EER for weighted sum rule fusion
    ( 2017-01-01)
    Nordiana Mukahar
    ;
    Haryati Jaafar
    ;
    Bakhtiar Affendi Rosdi
    In weighted sum rule fusion, proper weights assignment for different biometric sources is very important. The effective weights assignment procedure based on individual score performance is the vital key in adaptive weighted sum fusion. In this paper, an adaptive user specific weighted sum based on score distance is devised to determine optimal weights. It integrates the performance of the biometric source represented by its Equal Error Rate (EER) and score distance in weights estimation procedure. Although the designated weighted sum fusion requires less setting and less exhaustive search of fusion parameters, it can obtain better performance than previous approaches. The proposed technique outperforms the existing user specific weighted fusion schemes in all experiments by consistently achieves a minimum EER value.
  • Publication
    Image segmentation using k-means clustering and otsu's thresholding with classification method for human intestinal parasites
    ( 2020-07-09)
    Khairudin Norhanis Ayunie Ahmad
    ;
    Rohaizad Nurfatin Shamimi
    ;
    Aimi Salihah Abdul Nasir
    ;
    Lim Chee Chin
    ;
    Haryati Jaafar
    ;
    Mohamed Z.
    Helminth is one of the intestinal parasites that may cause harm and death to human. It is very important to have a system that is capable of assisting the technologist in investigating of fecal samples. In this paper, an automatic classification process is proposed to detect the different types of helminth eggs from fecal samples by using image processing technique. 50 samples of Ascaris Lumbricoides Ova (ALO) and Trichuris Trichiura Ova (TTO) are tested. First, these images undergo partial contrast stretching (PCS) technique to enhance the target images. Next, RGB and HSV color model have been compared in order to identify which color component is able to ease the segmentation process. S component shows a good results with high contrast between the target and the unwanted region. Then, Otsu's thresholding and k-means clustering are compared in order to to select the most suitable image processing method to be used in classification procedure. k-means clustering shows a better results compared to Otsu's thresholding. In classification process, area and size have been chosen as the feature to extract for the classification. The ratio for successfully detected ALO species is 84% while TTO is 76%.
      1
  • Publication
    Fast k-means clustering algorithm for malaria detection in thick blood smear
    ( 2020-11-09)
    Aris T.A.
    ;
    Aimi Salihah Abdul Nasir
    ;
    Lim Chee Chin
    ;
    Haryati Jaafar
    ;
    Mohamed Z.
    Lots of people all over the world is threaten by a popular blood infection illness that is called as malaria. According to this fact, immediate diagnosis tests are essential to avoid the malaria parasites from expanding in every part of the body. Malaria detection is based on parasitic count process on thick blood smear samples. Anyhow, this mechanism consist the chances of misinterpretation of parasites on behalf to human flaws. Thus, this research objective is to investigate the segmentation performance for improving malaria detection in thick blood smear images through fast k-means clustering algorithm on various color models. In this research, fast kmeans clustering is used because of its advantage which is no need to retrain cluster center that causes time taken to train the image cluster centers is reduce. Meanwhile, different color models have been utilized in order to identify the most relevant color model that obviously highlight the parasites. Five varied color models namely RGB, XYZ, HSV, YUV and CMY are selected and 15 color components namely R, G, B, X, Y, Z, H, S, V, Y, U, V, C, M and Y component have been derived with the aim to discover which color component is the topnotch for malaria parasites detection. In general, around 100 thick blood smear images have been tested in this study and the outcomes reveal that the best segmentation performance is segmentation through R component of RGB with 99.81% accuracy.
      1