Kekre's Fast Codebook Generation Research Articles

Age-Based Automatic Voice Conversion Using Blood Relation for Voice Impaired

The present work presents a statistical method to translate human voices across age groups, based on commonalities in voices of blood relations. The age-translated voices have been naturalized extracting the blood relation features e.g., pitch, duration, energy, using Mel Frequency Cepstrum Coefficients (MFCC), for social compatibility of the voice-impaired. The system has been demonstrated using standard English and an Indian language. The voice samples for resynthesis were derived from 12 families, with member ages ranging from 8–80 years. The voice-age translation, performed using the Pitch synchronous overlap and add (PSOLA) approach, by modulation of extracted voice features, was validated by perception test. The translated and resynthesized voices were correlated using Linde, Buzo, Gray (LBG), and Kekre’s Fast Codebook generation (KFCG) algorithms. For translated voice targets, a strong (θ >~93% and θ >~96%) correlation was found with blood relatives, whereas, a weak (θ

Content based Video Retrieval Systems Performance based on Multiple Features and Multiple Frames using SVM

In this paper, Content Based Video Retrieval Systems performance is analysed and compared for three different types of feature vectors. These types of features are generated using three different algorithms; Block Truncation Coding (BTC) extended for colors, Kekre’s Fast Codebook Generation (KFCG) algorithm and Gabor filters. The feature vectors are extracted from multiple frames instead of using only key frames or all frames from the videos. The performance of each type of feature is analysed by comparing the results obtained by two different techniques; Euclidean Distance and Support Vector Machine (SVM). Although a significant number of researchers have expressed dissatisfaction to use image as a query for video retrieval systems, the techniques and features used here provide enhanced and higher retrieval results while using images from the videos. Apart from higher efficiency, complexity has also been reduced as it is not required to find key frames for all the shots. The system is evaluated using a database of 1000 videos consisting of 20 different categories. Performance achieved using BTC features calculated from color components is compared with that achieved using Gabor features and with KFCG features. These performances are compared again with the performances obtained from systems using SVM and the systems without using SVM.

K-Means Codebook Optimization using KFCG Clustering Technique

Codebook Optimization is a concept of vector quantization which is applied to achieve lossy compression. Optimization of the codebook helps in maintaining the quality of the image. The codebook is generated using Kekre’s Fast Codebook Generation (KFCG) algorithm and Random Selection Method. The K-means algorithm is used to optimize the codebook. The Mean Square Error (MSE) is used as the measurement parameter. The point where the MSE converges is the optimal point and the codebook at that point is said to be the optimized codebook. The results obtained show that when K-means algorithm is applied to the codebook generated by KFCG Algorithm, less MSE is obtained as compared to Random Selection Method.

Codebook Design for Vector Quantization Based on a Kernel Fuzzy Learning Algorithm

Vector quantization (VQ) is an efficient technique for data compression and has been successfully used in various applications. The methods most commonly used to generate a codebook are the Linde, Buzo, Gray (LBG) algorithm, fuzzy vector quantization (FVQ) algorithm, Kekre‘s Fast Codebook Generation (KFCG) algorithm, discrete cosine transform based (DCT-based) codebook generation method, and k-principle component analysis (K-PCA) algorithm. However, if the separation boundaries in codebook generation are nonlinear, their performance can degrade fast. In this paper, we present a kernel fuzzy learning (KFL) algorithm, which takes advantages of the distance kernel trick and the gradient-based fuzzy clustering method, to create a codebook automatically. Experiments with real data show that the proposed algorithm is more efficient in its performance compared to that of the LBG, FVQ, KFCG, and DCT-based method, and to the K-PCA algorithm.

Comparative Analysis of Automatic Speaker Recognition using Kekre’s Fast Codebook Generation Algorithm in Time and Transform Domain

In this paper, an approach based on Kekre‟s fast code book Generation (KFCG) Algorithm in the transform domain has been proposed. KFCG is used for feature extraction in both the training and testing phases. Three methods for codebook generation have been used. In the 1 st method, codebooks are generated from the speech samples by using Discrete Fourier Transform (DFT). In the 2 nd method, the codebooks are generated using Discrete Cosine Transform (DCT). In the 3 rd method, the codebooks are generated using the Discrete Sine Transform (DST). For speaker identification, the codebook of the test sample is similarly generated and compared with the codebooks of the reference samples stored in the database. The results obtained for the above methods in the transform domain are compared with the results obtained in the time domain analysis. The results show that KFCG gives better results in transform domain than in time domain. Also the results improve as the vector dimension while generating the codebook is increased.

Vector Quantized Codebook Optimization Using Modified Genetic Algorithm

Vector Quantization (VQ) is lossy data compression technique and has various applications. Key to VQ is a good codebook. Once the codebook size is fixed, then the mean square error (MSE) reaches a value, beyond which it cannot be reduced by using codebook generation algorithms. In this paper, we are proposing modified genetic algorithm giving the optimal value, but it depends on the initial selection of the codevectors. Hence, it takes extremely huge time to give optimal value. For demonstration, we have used codebooks obtained from Linde Buzo and Gray (LBG) and Kekre’s Fast Codebook Generation (KFCG) algorithms. It is observed that the optimal error obtained from both LBG and KFCG is almost the same, indicating that they have converged to an optimal value. From the results, it is obvious that KFCG codebook gives slightly less minimized error as compared to LBG, indicating that KFCG codebook is closer to the optimum. The proposed method is general and can be applied to any clustering algorithm.

Performance Comparison of Speaker Recognition using Vector Quantization by LBG and KFCG

In this paper, two approaches for speaker Recognition based on Vector quantization are proposed and their performances are compared. Vector Quantization (VQ) is used for feature extraction in both the training and testing phases. Two methods for codebook generation have been used. In the 1 st method, codebooks are generated from the speech samples by using the Linde-Buzo-Gray (LBG) algorithm. In the 2 nd method, the codebooks are generated using the Kekre‟s Fast Codebook Generation (KFCG) algorithm. For speaker identification, the codebook of the test sample is similarly generated and compared with the codebooks of the reference samples stored in the database. The results obtained for both the methods have been compared. The results show that KFCG gives better results than LBG.