• Speech Sciences
• /
• v.12 no.3
• /
• pp.115-124
• /
• 2005

We investigated on the submission of multiple candidates in speaker-independent speech recognition by VQ/HMM. Submission of fixed number of multiple candidates has first been examined. As the number of candidates increases by two, three, and four, the recognition error rates were found to decrease by 41%, 58%, and 65%, respectively compared to that of a single candidate. We tried another approach that the candidates within a range of Viterbi scores are submitted. The number of candidates showed geometric increase as the admitted range becomes large. For a practical application, a combination of the above two methods was also studied. We chose the candidates within some range of Viterbi scores and limited the maximum number of candidates submitted to five. Experimental results showed that recognition error rates of less than 10% could be achieved with average number of candidates of 3.2 by this method.

• Journal of Korea Multimedia Society
• /
• v.7 no.12
• /
• pp.1737-1744
• /
• 2004

In the compression methods widely used today, the image compression by VQ is the most popular and shows a good data compression ratio. Almost all the methods by VQ use the LBG algorithm that reads the entire image several times and moves code vectors into optimal position in each step. This complexity of algorithm requires considerable amount of time to execute. To overcome this time consuming constraint, we propose an enhanced self-organizing neural network for color images. VQ is an image coding technique that shows high data compression ratio. In this study, we improved the competitive learning method by employing three methods for the generation of codebook. The results demonstrated that compression ratio by the proposed method was improved to a greater degree compared to the SOM in neural networks.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.27 no.8
• /
• pp.1265-1274
• /
• 1990

A new classified VQ scheme for DCT coefficients(DCT-CVQ) is proposed for still image coding. DCT coefficient matrices are classified into six classes by ac coefficients that well represent edge characteristics and the distribution of ac energy in the DCT domain. To reduce the complexity of, VQ, ac transform coefficients are partitioned into several vectors, and an adaptive method is provided by making different codebook sizes for different classes. Simulation results show that this DCT-CVQ results in good subjective quality at low bit rates, and SNR is increased by about 1.5-4.0 dB in the range 0.2-0.5 bpp, when compared with other DCT-VQ coding schemes.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.18 no.1
• /
• pp.47-56
• /
• 1993

In this paper, pyramidal VQ image coding by DCT classifier based on HVS is studied. The proposed DCT classifier based on HVS is that the transform subblocks of the image are mlultiplied by MTF which is a sort of band pass filter and sorted by the magnitude of their ac energy levels and classifeid into three classes such as low, middle and high variance class by the threshold and then edges are detected in comparison of the energy sum of ac transform coefficients corresponding to the different edge directions.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.3
• /
• pp.57-65
• /
• 1995

This paper presents a new classified vector quantization (VQ) technique using a neural network model in the transform domain. Prior to designing a codebook, the proposed approach extracts class features from a set of images using self-organizing feature map (SOFM) that has the pattern recognition characteristics and the same as VQ objective. Since we extract the class features from the training images unlike previous approaches, the reconstructed image quality is improved. Moreover, exploiting the adaptivity of the neural network model makes our approach be easily applied to designing a new vector quantizer when the processed image characteristics are changed. After the generalized BFOS algorithm allocates the given bits to each class, codebooks of each class are also generated using SOFM for the maximal reconstructed image quality. In experimental results using monochromatic images, we obtained a good visual quality in the reconstructed image. Also, PSNR is comparable to that of other classified VQ technique and is higher than that of JPEG baseline system.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.27 no.9
• /
• pp.1468-1475
• /
• 1990

In this paper, isolated words are recognized using multisection VQ and HMM. As recognition vocabuaries, 20 area-name which is uttered 5 times by 3 speakers is selected. In generating codebook, we devide recognition vocabulary into equal length, section, and make standard VQ codebook to each section and calculate observation by section and than recognize isolated words by HMM training. Multisection VQ codebook has time information and as observation is calculated by eacy section, computation is lesser and recongnition rate is higher than by whole codword. As a result, it is proved that recognition rate is higher in case of HMM using multisection VQ codebook.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2000.10b
• /
• pp.883-886
• /
• 2000

Two fast search algorithms are proposed for VQ encoding in subband/VQ coding schemes. These algorithms exploit the property of the transform domain that the large coefficients rarely exist in the decomposed subbands. And the exit condition of PDE algorithm can be satisfied by comparing the large values of the codeword with the corresponding ones of the input vector. The computational complexity can be reduced at the expense of memories without extra coding errors.

• The Journal of the Acoustical Society of Korea
• /
• v.16 no.2
• /
• pp.95-100
• /
• 1997

In this paper, we present a new fast VQ encoding algorithm. The proposed algorithm facilitates two characteristics of a vector, i.e., mean and variance to reject many unlikely codewords and save a lot of computation time. Since the proposed algorithm, which is based upon geometric considerations, rejects those codewords that are impossible to be the closest codeword, it provides the same results as a conventional exhaustive(or full) search algorithm. The simulation results confirm the effectiveness of the proposed algorithm.

• Proceedings of the KIEE Conference
• /
• 2001.07d
• /
• pp.2634-2636
• /
• 2001

This paper proposes a new method of retrieving images from large image databases. The method is based on VQ(Vector Quantization) of local texture information at interest points automatically detected in an image. The texture features are extracted by Gabor wavelet filter bank, and rearranged for rotation. These features are classified by VQ and then construct a pattern histogram. Retrievals are performed by just comparing pattern histograms between images. Experimental results have shown the robustness of the proposed method to image rotation, small scale change, noise addition and brightness change and also shown the possibility of the retrieval by a partial image.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.25 no.9B
• /
• pp.1596-1603
• /
• 2000

In this paper an image coding scheme using the correlation of neighboring VQ indices in wavelet domain is proposed. The codewords in each sub-codebook are reordered in ascending order according to their energy in order to inmcrease the dependency of the indices. And the relative dresses of the reordered indices are transmitted to a receiver instead of to absolute addresses. LBG algorithm is used for obtaining a multiresolution codebool and PNN using k-d tress used for generating its initial values. Our experimental results show that the proposed coding scheme outperforms JPEG and an ordinary wavelet-VQ method at low-bit rates in terms of PSNR.