Kangeun Lee

An efficient LSF quantization using dynamic bit allocation

The current paper is concerned with an effective method to quantize a spectrum envelope of a speech signal without having an inter-frame prediction. In this paper, we proposed a method referred to as dynamic bit allocation-split vector quantization (DBA-SVQ). The main feature of this structure is that it makes use of the ordering property of line spectral frequencies (LSF) and exploits multiple codebooks, normalization and the DBA technique. As a result, we can limit the dynamic range of LSF sub-vectors and allocate different numbers of bits in accordance with the range sizes to maximize the overall efficiency of quantization. The performance is compared with delta line spectral pairs (LSP) VQ, which is used in EVRC-B, demonstrating reduction in spectral distortion (SD).

Embedded real-time S/W beamforming platform with reconfigurable multi-core processors

In modern medical ultrasound signal processing, beamforming is usually implemented on H/W solution by ASIC(Application Specific Integrated Circuit) because of its huge computation. ASIC solution has a problem with flexibility to support various beamforming algorithms. Nowadays, computing ability of GPU(Graphic processing unit) becomes very high, therefore many approaches have been proposed for S/W beamforming on GPU. Although the high performance of GPU, commercial GPU is not proper for portable ultrasound, because of its large power consumption. The motivation of this paper is evaluating the feasibility of embedded multi-core system as S/W beamforming solution for portable ultrasound. To develop embedded S/W beamforming platform, we propose the platform with multiple embedded processors called RP(Reconfigurable Processor) and special co-processors. Whole system is composed of 6 computing clusters and single cluster is composed of 8 RP processors and 1 co-processor. The number of clusters in the system can be changed depending on computational requirement. To evaluate the performance of the proposed platform, we implemented MV(Minimum Variance) beamforming, which is one of the most complex beamforming, on that platform. 4 approaches were mainly used to accelerate MV beamforming. The first one is co-processor for accelerating MAC(Multiply and Accumulate) operations, the second one is special instructions for beamforming, the third one is SIMD(Single Instruction Multiple Data), and the last one is CGA(Corse Grained Architecture) acceleration which is special function of RP. As a final result, 128channel 30fps(frame per second) real-time MV beamformer was achieved on the proposed platform.

Virtual reproduction of spherical multichannel sound over 5.1 speaker system

A virtual sound rendering is proposed that allows conventional 5.1 speaker system to reproduce spherical sound including vertical direction. The proposed algorithm is based on vector based amplitude panning (VBAP) and uses a perceptual model of human. The virtual spherical sound was compared with 5.1 sound and demonstrated significantly enhanced spatiality.