Hasil Pencarian  ::  Simpan CSV :: Kembali

"ABSTRACT

This study describes bispectrum pattern analysis and quantization for identifying speaker in noisy environment. Direct, non-parametric, bispectrum analysis and estimation was performed before quantization and classification process. As for reliable quantization approach this study applied an algorithm of vector quantization method using combined Self Organizing Feature Map (SOFM) and Learning Vector Quantization (LVQ) neural network, to quantize bispectrum of speech data. Since there is no prior knowledge on bispectrum data distribution to determine class information, we used an adaptive codebook generation method, which is a hybrid of SOFM to generate the codebook internally and LVQ algorithm to improve the cluster distribution in the classifier decision. In addition with the SOFM+LVQ algorithm, a nonlinear vector quantization method (NLVQ) is introduced in dealing with a case where there is a low-separability problem of codebook data obtained from one speaker. This new NLVQ technique employs a nonlinear third order hyperbolic tangent function which combines noise suppression effect with a dynamic range limitation, in or-der to transform the bispectrum input data to be used for making the codebook. Nearest-neighbor rule statistical analysis was used to estimate the recognition performance of the system before classification."

"Penelitian ini melakukan studi mengenai sistem identifikasi pembicara berbahasa Indonesia menggunakan SVM. Parameter sistem terdiri atas silence removal, PCA, nilai rata-rata dan varians MFCC. Ujicoba menggunakan data berita berbahasa Indonesia dari televisi dan radio yang disegmen dalam 5, 10, 15 detik dengan jumlah data 26 jam (715 pembicara). Hasil penelitian ini menunjukkan ketepatan pengenalan pembicara sebesar 94-98% untuk kombinasi parameter silence removal dan rata-rata MFCC dengan akurasi terbaik pada segmen waktu 10 detik. Namun dengan bertambahnya jumlah pembicara, ketepatan pengenalan cenderung berkurang. Penelitian ini dapat dikembangkan untuk sistem perolehan informasi data speech berdasarkan siapa yang berbicara dalam suatu sesi data.

---

This research studies speaker identification system for Indonesian speech based on SVM. Parameters of this system are silence removal, PCA, average and varians values of MFCC. The experiments use 26 hours (715 speakers) Indonesian broadcast news from radio and television segmented into 5, 10, 15 seconds. The results achieve 94-98% identification accuracy for combination of parameters silence removal and average of MFCC. The best accuracy comes from 10 seconds time segment. However, the accuracy falls when the number of speakers increases. This study could be used for speech retrieval system based on who speaks in a speech session."

"This book examines use of the voice as a biometric measure for personal authentication, offering an overview of advances in speaker authentication, and including useful algorithms and techniques for improving overall system robustness and performance."

"Industri ritel di Indonesia mengalami pertumbuhan yang pesat setiap tahunnya. Pertumbuhan tersebut membuat tingkat kompetisi dan tantangan pada industri ini semakin tinggi. Salah satu tantangan yang harus dihadapi antara lain perubahan pola belanja masyarakat. Dalam memahami perilaku pola belanja masyarakat, pencarian pola asosiasi antar produk yang dibeli konsumen dapat dilakukan dengan menggunakan metode Market Basket Analysis. Dua algortima yang digunakan yaitu algoritma Apriori dan Speaker-listener Label Propagation Algorithm SLPA. Dari hasil aturan yang didapat berdasarkan algoritma Apriori diperoleh tiga belas aturan dan berdasarkan hasil algoritma SLPA diperoleh 41 komunitas dari kategori produk sebagai luaran akhir penelitian.

---

The retail industry in Indonesia has been rapidly growing each year. The growth and development of retail industry in Indonesia also causes higher level of competition and challenge within the scope of industrial environment. One of the challenges is the pattern of public spending. In order to understand this challenge, it is necessary to conduct a research about pattern between products purchased by the customers which can be done using Market Basket Analysis Method. There are two algorithms used in this research, which are Apriori algorithm and Speaker listener Label Propagation Algorithm SLPA. In the result, Apriori algorithm obtained 13 rules whilst SLPA algorithm resulted in 41 communities from the product category as the final output of the research."

"Indonesia has the world's largest tropical peatlands of about 14.9 million hectares that have important life support roles. However, fire frequently occurs in peatlands. According to experts and field forest fire fighters, fire hotspots that appear in a sequence of two to three days at the same location has a high potential of becoming a forest fire. This study aimed to determine the sequential patterns of hotspot occurrences, classify satellite image data and identify the fire spots. Fire spot identification was done using hotspot sequence patterns that were overlaid with burned area classification results. Sequential pattern mining using the Prefix Span algorithm was applied to identify sequences of hotspot occurrence. Maximum Likelihood method was applied to classify Landsat 7 satellite images toward identifying burned areas in Pulang Pisau and Palangkaraya in Central Kalimantan and Pontianak in West Kalimantan. Sequence patterns were overlaid with image classification results. The study results show that in Pulang Pisau, 26.19% of sequence patterns are located in burned areas and 72.62% sequence patterns were found in the buffer of burned area within a radius of one kilometer. As for Palangkaraya, there were 62.50% sequence patterns located in burned areas and 87.50% sequence patterns in the buffer of burned area with the radius of one kilometer. In total, Â there were 72.62% and 87.50% fire hotspots recorded in Pisau and Palangkaraya, respectively which are strong indicators of peatland fires."

"ABSTRAKDistribusi normal adalah salah satu jenis persebaran kelompok data yang didefinisikan berdasarkan rata-rata dan standar deviasi dari sekelompok data, yang dapat digunakan untuk mengelompokkan data berdasarkan posisinya terhadap standar deviasi dari kelompok data tersebut. Learning Vector Quantization adalah salah satu jenis neural network yang bisa mempelajari sendiri masukan yang ia terima kemudian memberi keluaran sesuai dengan masukan tersebut, dengan metode supervised dan competitive learning. Skripsi ini membahas penerapan dan analisis dari kedua sistem tersebut untuk menguji hasil deteksi plagiarisme oleh sistem deteksi plagiarisme berbasis latent semantic analysis, yang berasal dari program Simple-O. Beberapa modifikasi dilakukan untuk meningkatkan akurasi pengujian, antara lain dengan melakukan variasi parameter-parameter dari metode distribusi normal, yakni dengan mengubah batas standar deviasi maupun dengan mengubah koefisien pengali batas nilai pada standar deviasi tertentu, dimana hasilnya adalah standar deviasi maupun koefisien pengalinya berbanding lurus dengan aspek relevansi program (recall) namun tidak pada akurasi (F-Measure). Modifikasi juga dilakukan pada parameter percepatan belajar dari algoritma learning vector quantization, dimana hasilnya adalah parameter percepatan belajar berbanding terbalik dengan relevansi program maupun akurasi. Kemudian variasi dan analisis dilakukan pada tujuh jenis besaran hasil keluaran sistem deteksi plagiarisme berbasis latent semantic analysis, yakni frobenius norm, slice, dan pad, beserta kombinasinya, dimana hasilnya keberadaan frobenius norm diwajibkan untuk melakukan evaluasi kemiripan antara dua teks. Kemudian hasil pengujian menggunakan kedua metode digabungkan menggunakan operasi AND yang memberikan hasil yang beragam, dengan catatan perlunya keseimbangan antara precision dan recall dari masing pengujian yang akan dilakukan operasi AND untuk memberikan hasil yang baik. Dengan menggunakan kombinasi metode dan parameter yang tepat, terdapat peningkatan akurasi sistem dari 35-46% pada penelitian sebelumnya hingga maksimal 65,98%.

---

ABSTRACT

Normal distribution is a type of data distributions which is defined from the average and standard deviation of the data cluster. It can be used to group datas based on its position from the standard deviation of the data cluster. Learning vector quantization is a type of neural networks that can learn from inputs it gets to give appropriate outputs, with supervised and competitive learning methods. This thesis discusses the implementation and analysis of both methods to verify the plagiarism detection results from detection plagiarism system based on latent semantic analysis, which is based on Simple-O program. Some modifications are made, such as by variating the parameters of normal distribution method, by changing the limits of standard deviation or by changing the factor of the number limit at a particular standard deviation. Both of them appear to be directly proportional to the relevance (recall), but not with accuracy (F-Measure). Modifications are also made at the learning acceleration parameters from the learning vector quantization algorithm, which sees the parameters being inversely proportional to both the relevance and accuracy. Then, variations and analysis are done to seven types of magnitude from the results of the plagiarism detection system, which are frobenius norm, slice, and pad, and their combinations, which suggest that frobenius norm is the most verifiable results, and must be included to be evaluated when text similarity analysis are conducted. Then, verification results using both methods are combined using AND operation which gives diverse results. However, it is needed to have a balance between precision and recall from each verifications to produce good results. With correct combinations of methods and parameters, system accuracy are increased from 35-46% of last research to maximum accuracy of 65,98%."