公开(公告)号：KR1020130037041A

公开(公告)日：2013-04-15

申请号：KR1020110101360

申请日：2011-10-05

Applicant: 목포대학교산학협력단

Inventor： 이성로 ,  이연우 ,  이정훈 ,  김한수 ,  김동진 ,  장준혁 ,  김남수 ,  김홍국

IPC: G10L15/20 ,  G10L15/14 ,  G10L19/00 ,  G10L21/02

CPC classification number: G10L15/14 ,  G10L19/038 ,  G10L21/02

Abstract: PURPOSE: A sound model compensation method which combines mask based multi-channel sound separation for marine sound is provided to analyze noise and voice properties, thereby improving performance of voice recognition. CONSTITUTION: A noise model and a weighted value necessary for sound model adaptation are obtained by using an HMM based mask. Mask information is estimated from a binaural signal. Noise is decreased by using the mask. After MMSS(Mask-based Multi-channel Source Separation) is performed, a 13th MFCC(Mel Frequency Cepstral Coefficient) is extracted from a voice signal.

Abstract translation: 目的：提供一种组合基于屏蔽的多声道声音分离用于海洋声音的声音模型补偿方法，用于分析噪声和声音特性，从而提高语音识别的性能。 构成：通过使用基于HMM的掩模获得声音模型适应所需的噪声模型和加权值。 屏蔽信息由双耳信号估计。 使用面罩降低噪音。 在执行MMSS（基于掩码的多信道源分离）之后，从语音信号中提取第13个MFCC（Mel频率倒谱系数）。

公开(公告)号：KR1020130034915A

公开(公告)日：2013-04-08

申请号：KR1020110099080

申请日：2011-09-29

Applicant: 목포대학교산학협력단

Inventor： 임정수 ,  장준혁 ,  김홍국 ,  김남수 ,  박선 ,  박희만 ,  정민아 ,  이연우 ,  이성로

IPC: G10L21/02 ,  G10L19/00

Abstract: PURPOSE: A method for efficiently reinforcing a voice signal based on a voice coding parameter in marine voice communications is provided to reinforce excitation signals among a bitstream parameter based on a voice encoding device in root environment, thereby supplying excellent performance. CONSTITUTION: An excitation signal gain ratio estimation unit(140) estimates an excitation signal gain ratio by using an excitation signal gain of background noises estimated by an excitation signal gain estimation block(130), and a decoding result which applies a fixed codebook gain by a fixed codebook vector decoding block(120). An excitation signal generation unit(150) generates an excitation signal by using a decoding result which applies an adaptive codebook gain by an excitation signal gain ratio estimation result and an adaptive codebook vector decoding block(110). [Reference numerals] (110,120) Adaptive codebook vector decoding; (130) Excitation signal gain estimation of background noise; (140) Excitation signal gain ratio estimation; (150) Excitation signal generation; (160) Linear prediction synthesis filter; (170) Postprocessor; (AA,BB) Input bit stream; (CC) Adaptive codebook gain application; (DD) Fixed codebook gain application; (EE) Enhanced voice signal; (FF) Near-end background noise estimation;

Abstract translation: 目的：提供一种在海洋语音通信中基于语音编码参数有效加强语音信号的方法，以加强基于根环境中的语音编码装置的比特流参数中的激励信号，从而提供出色的性能。 激励信号增益比估计单元（140）通过使用由激励信号增益估计块（130）估计的背景噪声的激励信号增益来估计激励信号增益比，以及解码结果，其通过 固定码本向量解码块（120）。 励磁信号生成单元（150）通过使用通过激励信号增益比估计结果应用自适应码本增益的解码结果和自适应码本矢量解码块（110）来产生激励信号。 （110,120）自适应码本矢量解码; （130）背景噪声的激励信号增益估计; （140）励磁信号增益比估计; （150）励磁信号发生; （160）线性预测合成滤波器; （170）后处理; （AA，BB）输入位流; （CC）自适应码本增益应用; （DD）固定码本增益申请; （EE）增强语音信号; （FF）近端背景噪声估计;

公开(公告)号：KR1020120048363A

公开(公告)日：2012-05-15

申请号：KR1020100109962

申请日：2010-11-05

Applicant: 목포대학교산학협력단

Inventor： 정지원 ,  김민혁 ,  박태두 ,  김남수 ,  이성로

IPC: H04N19/89 ,  H04N7/015

CPC classification number: H03M13/1125 ,  H03M13/1165

Abstract: PURPOSE: A decoding method using a likelihood ratio is provided to selectively delete a symbol with a high error probability according to an LLR(Log Likelihood Ratio) instead of deleting the whole Internet protocol packet. CONSTITUTION: An IP(Internet Protocol) packet receives a LDPC(Low-Density Parity-Check) encoded signal. The encoded LDPC signal is decoded by the LDPC code. A symbol unit deletion in the IP packet is determined by an outputted LLR. An IP packet deleted by the symbol unit is stored in a memory. An error correction decoding is performed.

Abstract translation: 目的：提供使用似然比的解码方法，以根据LLR（对数似然比）来选择性地删除具有高错误概率的符号，而不是删除整个因特网协议分组。 构成：IP（因特网协议）分组接收LDPC（低密度奇偶校验）编码信号。 编码的LDPC信号由LDPC码解码。 IP分组中的符号单元删除由输出的LLR确定。 由符号单元删除的IP包被存储在存储器中。 执行纠错解码。

公开(公告)号：KR101021481B1

公开(公告)日：2011-03-16

申请号：KR1020090105819

申请日：2009-11-04

Applicant: 목포대학교산학협력단

Inventor： 조정용 ,  박근형 ,  김남수 ,  문제학 ,  김선재 ,  마승진 ,  김향선

IPC: C12G1/022

CPC classification number: C12G1/0203 ,  C12G2200/05 ,  C12G2200/21

Abstract: PURPOSE: A method for producing green tea white wine is provided to ensure functionality and to improve preference under optimal condition. CONSTITUTION: A method for producing green tea white wine comprises: a step of crushing and pressing green grapes to obtain grape juice; a step of inoculating fermentation strain to the grape juice; a step of fermenting the grape juice to obtain a primary fermentation; a step of adding dried green tea into the primary fermentation; and a step of secondarily fermenting. The fermentation strain is Saccharomyces cerevisiae KCCM 11351, Saccharomyces cerevisiae KCCM 12224, Saccharomyces bayanus KCCM 12633, and Saccharomyces pastorianus KCCM 32361.

Abstract translation: 目的：提供一种生产绿茶白葡萄酒的方法，以确保最佳条件下的功能和改善偏好。 构成：生产绿茶白葡萄酒的方法包括：将绿色葡萄粉碎和压榨获得葡萄汁的步骤; 向葡萄汁接种发酵菌株的步骤; 发酵葡萄汁以获得初级发酵的步骤; 将干绿茶加入初级发酵的步骤; 和二次发酵的步骤。 发酵菌株是酿酒酵母KCCM11351，酿酒酵母KCCM 12224，酿酒酵母KCCM 12633和巴斯德毕赤酵母KCCM 32361。