• Proceedings of the IEEK Conference
• /
• summer
• /
• pp.391-394
• /
• 2004

This paper focuses on the DSP implementation of an HMM-based speech recognizer that can handle several hundred words of vocabulary size as well as speaker independency. First, we develop an HMM-based speech recognition system on the PC that operates on the frame basis with parallel processing of feature extraction and Viterbi decoding to make the processing delay as small as possible. Many techniques such as linear discriminant analysis, state-based Gaussian selection, and phonetic tied mixture model are employed for reduction of computational burden and memory size. The system is then properly optimized and compiled on the TMS320C6711 DSP for real-time operation. The implemented system uses 486kbytes of memory for data and acoustic models, and 24.5kbytes for program code. Maximum required time of 29.2ms for processing a frame of 32ms of speech validates real-time operation of the implemented system.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.33 no.10A
• /
• pp.997-1005
• /
• 2008

In this paper, we propose an efficient space-time block coding (STBC) scheme in a cooperative satellite-terrestrial system. The proposed STBC scheme has code rate 1 for a 3 transmit antenna scheme. Because the channel matrix of the proposed scheme is orthogonal, we can use a simple linear decoding algorithm and also can expect improved performance over the conventional scheme. The simulation results demonstrate that the proposed scheme has improved performance for bit error rates (BER) than several conventional STBC schemes. In addition, we investigate performance simulation results by power imbalance between the terrestrial repeaters and satellite.

• Journal of Broadcast Engineering
• /
• v.13 no.4
• /
• pp.452-462
• /
• 2008

In this paper, we address issues associated with the real-time implementation of the MPEG-1/2 Layer-II (or MUSICAM) and MPEG-4 ER-BSAC decoders for Digital Multimedia Broadcasting (DMB) on TMS320C64x+ that is a fixed-point DSP processor with a clock speed of 330 MHz. To achieve the real-time requirement, they should be optimized in different steps as follows. First of all, a C-code level optimization is performed by sharing the memory, adjusting data types, and unrolling loops. Next, an algorithm level optimization is carried out such as the reconfiguration of bitstream reading, the modification of synthesis filtering, and the rearrangement of the window coefficients for synthesis filtering. In addition, the C-code of a synthesis filtering module of the MPEG-1/2 Layer-II decoder is rewritten by using the linear assembly programming technique. This is because the synthesis filtering module requires the most processing time among all processing modules of the decoder. In order to show how the real-time implementation works, we obtain the percentage of the processing time for decoding and calculate a RMS value between the decoded audio signals by the reference MPEG decoder and its DSP version implemented in this paper. As a result, it is shown that the percentages of the processing time for the MPEG-1/2 Layer-II and MPEG-4 ER-BSAC decoders occupy less than 3% and 11% of the DSP clock cycles, respectively, and the RMS values of the MPEG-1/2 Layer-II and MPEG-4 ER-BSAC decoders implemented in this paper all satisfy the criterion of -77.01 dB which is defined by the MPEG standards.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.38C no.10
• /
• pp.858-866
• /
• 2013

In the DVB-S2, which is the technical specification of the second generation digital video broadcasting via satellite, APSK modulation scheme along with LDPC coding schemes are defined. APSK is a multi-lelvel PSK modulation scheme and decoding of LDPC coded signal requires soft decision information. Therefore, the APSK demodulator at the receiver should have capability of estimating soft information. In this paper, we introduce a method to estimate soft information by using simple distance estimation, and show that this method overestimates the soft information. Subsequently, this overestimated soft information leads to performance degradation. In order to overcome this problem, we propose a scaling method to improve the performance at the receiver In addition, we show that the proposed scaling scheme enables us to estimate the soft information with linear order complexity and produce the performance close to the maximum likelihood detection.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.27 no.10A
• /
• pp.1021-1028
• /
• 2002

This paper proposes a method to extract the content curve that reflects changes in video content from the MPEG video in the compressed domain, and also describes a video summary generation technique which can read video effectively and rapidly from the content curve. Existing video summary techniques have a disadvantage of taking significant amount of time to generate the video summary due to complex calculations in the decoding process. Moreover, the existing techniques, which use video content curve, require to perform many calculations to process the high dimensional content curve. However, the proposed technique generates video summary fast via the linear approximation of the proposed curve, after extraction the two dimensional content curve directly. In addition, the proposed technique has a merit that the user can set any number of key-frames and amount of calculation that form the video summary.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.38A no.4
• /
• pp.295-302
• /
• 2013

In the network coding, throughput can be increased by allowing the transformation of the received data at the intermediate nodes. However, the adversary can obtain more information at the intermediate nodes and make troubles for decoding of transmitted data at the sink nodes by modifying transmitted data at the compromised nodes. In order to resist the adversary activities, various information theoretic or cryptographic secure network coding schemes are proposed. Recently, a secure network coding based on the cryptographic hash function can be used at the random network coding. However, because of the computational resource requirement for cryptographic hash functions, networks with limited computational resources such as sensor nodes have difficulties to use the cryptographic solution. In this paper, we propose a new secure network coding scheme which uses linear transformations and table lookup and safely transmits n-1 packets at the random network coding under the assumption that the adversary can eavesdrop at most n-1 nodes. It is shown that the proposed scheme is an all-or-nothing transform (AONT) and weakly secure network coding in the information theory.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.12 no.7
• /
• pp.1235-1242
• /
• 2008

In this paper, x-ray image sensor of photon counting type having a $32{\times}32$ pixel array is designed with $0.18{\mu}m$ triple-well CMOS process. Each pixel of the designed image sensor has an area of loot $100{\times}100\;{\mu}m2$ and is composed of about 400 transistors. It has an open pad of an area of $50{\times}50{\mu}m2$ of CSA(charge Sensitive Amplifier) with x-ray detector through a bump bonding. To reduce layout size, self-biased folded cascode CMOS OP amp is used instead of folded cascode OP amp with voltage bias circuit at each single-pixel CSA, and 15-bit LFSR(Linear Feedback Shift Register) counter clock generator is proposed to remove short pulse which occurs from the clock before and after it enters the counting mode. And it is designed that sensor data can be read out of the sensor column by column using a column address decoder to reduce the maximum current of the CMOS x-ray image sensor in the readout mode.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.5
• /
• pp.197-206
• /
• 2014

Motor imagery based Brain-computer Interface(BCI), which has recently attracted attention, is the technique for decoding the user's voluntary motor intention using Electroencephalography(EEG). For classifying the motor imagery, event-related desynchronization(ERD), which is the phenomenon of EEG voltage drop at sensorimotor area in ${\mu}$-band(8-13Hz), has been generally used but this method are not free from the performance degradation of the BCI system because EEG has low spatial resolution and shows different ERD-appearing band according to users. Common spatial pattern(CSP) was proposed to solve the low spatial resolution problem but it has a disadvantage of being very sensitive to frequency-band selection. Discriminative filter bank common spatial pattern(DFBCSP) tried to solve the frequency-band selection problem by using the Fisher ratio of the averaged EEG signal power and establishing discriminative filter bank(DFB) which only includes the feature frequency-band. However, we found that DFB might not include the proper filters showing the spatial pattern of ERD. To solve this problem, we apply a band-selection process using CSP feature vectors and linear discriminant analysis to DFBCSP instead of the averaged EEG signal power. The filter selection results and the classification accuracies of the existing and the proposed methods show that the CSP feature is more effective than signal power feature.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.28 no.5C
• /
• pp.532-538
• /
• 2003

In this paper we have investigated a novel approach applying low-density parity-check coding to a COFDM-CDMA system, which operates in a multi-path fading mobile channel. Developed as a linear-block channel coder, the LDPC code is known for a superior signal reception capability in AWGN and/or flat fading channels with respect to increased encoding rates, however, its performance degrades when the communication channel becomes multi-path fading. For a typical multi-path fading mobile channel with a SNR of 16㏈ or lower. in order to obtain a BER lower than 1 out of 10000, the LDPC code with encoding rates below 1:3 requires not only the inherent parity check information but also the piloting information for refreshing front-end equalizer taps of COFDM-CDMA, periodically. For instance, while the 1:3-rate LDPC coded transmission symbol is consisted of data bits and parity-check bits in 1 to 3 proportion, on the other hand, in the proposed method the same rate LDPC transmission symbol contains data bits, parity check bits, and pilot bits in 1 to 2 to 1 proportion, respectively. The included pilot bits are effective not only for channel estimation and channel equalization but for symbol decoding by assisting the parity-check bits, hence, improving SNR vs BER performance over the conventional 1:3-rate LDPC code. The proposed system performance has been verified using computer simulations in multi-path, Rayleigh fading channels, and the results show us that the proposed method out-performs the general LDPC channel coding methods in terms of SNR vs BER measurements.