-
公开(公告)号：KR1020150054424A
公开(公告)日：2015-05-20
申请号：KR1020130136846
申请日：2013-11-12
Applicant: 고려대학교 산학협력단
IPC: G10L21/02
CPC classification number: G10L21/02 , G10L21/0208
Abstract: 잔여에코제거방법및 그장치가개시된다. 잔여에코제거방법은 (a) 입력신호의상태를검출하는단계; (b) 상기검출된상태에따라필터가중치갱신여부를결정하여상기입력신호에대한 1차에코를제거하는단계; (c) 스피커를통해출력되는후단신호(far-end signal)를분류하는단계; 및 (d) 상기분류결과에따라상기 1차에코제거된신호에서파워스펙트럼밀도추정에따른하모닉성분을이용한잔여에코성분추정여부를결정하여잔여에코를제거하는단계를포함한다.
Abstract translation: 提供了一种残余回波消除方法及其装置。 剩余回声消除方法包括以下步骤：（a）检测输入信号的状态; （b）通过参考检测到的状态来决定滤波器加权的更新来取消输入信号的第一回波; （c）对通过扬声器输出的远端信号进行分类; 以及（d）基于根据分类结果取消回波的第一信号的功率谱的密度估计，通过使用谐波分量来决定残留回波的分量估计来消除残差回波。
-
公开(公告)号：KR1020140121169A
公开(公告)日：2014-10-15
申请号：KR1020130037589
申请日：2013-04-05
Applicant: 고려대학교 산학협력단
CPC classification number: G10L15/26 , G10L15/02 , G10L15/18 , G10L21/06 , G10L2021/065
Abstract: 본 발명은 청각장애인용 상황대처 음성 자막 변환 장치 및 방법을 개시한다. 본 발명에 따르면, 복수의 상황별 음성 모델 및 핵심어 모델을 저장하는 상황 데이터베이스; 사용자가 선택한 상황에 상응하는 음성 모델을 참조하여 마이크를 통해 입력된 음성 신호를 인식하여 음성 인식 결과를 출력하는 음성 인식부; 및 사용자가 선택한 상황에 상응하는 핵심어 모델을 참조하여 상기 음성 인식 결과에서 핵심어를 추출하는 핵심어 추출부를 포함하는 청각장애인용 음성 자막 변환 장치가 제공된다.
Abstract translation: 公开了一种用于听障人士的情况自适应语音文本转换装置及其方法。 根据本发明的用于听障人士的情况自适应语音到文本转换装置包括：存储用于多种情况的语音模型和关键词模型的情况数据库; 语音识别部，其通过参照与用户选择的情况相对应的语音模型识别通过麦克风输入的语音信号，并输出语音识别结果; 以及关键字提取部，通过参照与用户选择的情况对应的关键字模型，从语音识别结果中提取关键字。
-
公开(公告)号：KR101412369B1
公开(公告)日：2014-07-02
申请号：KR1020130011907
申请日：2013-02-01
Applicant: 고려대학교 산학협력단
CPC classification number: G06T7/44 , G06T7/11 , G06T7/254 , G06T2207/30201
Abstract: The present invention relates to a method and an apparatus for detecting an object from an image regardless of the rotational change thereof. A method for detecting an object from an image according to an embodiment of the present invention comprises the steps of: maintaining a database in which a first characteristic value of a first object image and a second characteristic value of a second object image extracted by using a histogram extracted from a sinogram of an image are divided and stored; receiving an input image; performing the Radon transform of the input image, and extracting a characteristic value of the input image by extracting a histogram from the Radon transform result; and detecting a specific object from the input image by comparing the characteristic value of the input image with the first and second characteristic values stored in the database.
Abstract translation: 本发明涉及一种用于从图像中检测物体的方法和装置，而不管其旋转变化如何。 根据本发明的实施例的用于从图像检测对象的方法包括以下步骤：维护数据库，其中通过使用第一对象图像提取第一对象图像的第一特征值和第二对象图像的第二特征值 从图像的正弦图提取的直方图被分割和存储; 接收输入图像; 执行输入图像的Radon变换，以及通过从Radon变换结果提取直方图来提取输入图像的特征值; 以及通过将输入图像的特征值与存储在数据库中的第一和第二特征值进行比较来检测来自输入图像的特定对象。
-
公开(公告)号：KR1020140033262A
公开(公告)日：2014-03-18
申请号：KR1020120093894
申请日：2012-08-27
Applicant: 한화테크윈 주식회사 , 고려대학교 산학협력단
CPC classification number: H04N7/18 , G06K9/00771 , G06K9/6277 , G06T7/11 , G06T7/194 , G06T2207/30232
Abstract: Disclosed by the present invention are an indoor surveillance system and an indoor surveillance method. The indoor surveillance system of the present invention includes: a sound processor which extracts a sound feature from a sound signal and determines whether an abnormal event occurs in a surveillance area based on the extracted sound feature; an image processor which detects a foreground blob from an image signal and determines whether an abnormal event occurs in the surveillance area based on motion information of the detected foreground blob; and a context-aware device which produces a sound abnormality probability and an image abnormality probability by accumulating an abnormal event determination result about the sound signal and an abnormal event determination result about the image signal over a fixed time period and finally determines whether the abnormal event occurs in the surveillance area by using a combination abnormal probability distribution model about a normal situation and an abnormal situation. [Reference numerals] (20) Sound sensor; (30) Image sensor; (40) Input device; (50) Sound processor; (60) Image processor; (70) Context-aware device; (80) Output device
Abstract translation: 本发明公开了一种室内监视系统和室内监视方法。 本发明的室内监视系统包括：声音处理器，其从声音信号中提取声音特征，并且基于所提取的声音特征来确定在监视区域中是否发生异常事件; 图像处理器，其从图像信号检测前景斑点，并且基于检测到的前景斑点的运动信息来确定在监视区域中是否发生异常事件; 以及通过在固定时间段内累积关于声音信号的异常事件确定结果和关于图像信号的异常事件确定结果来产生声音异常概率和图像异常概率的上下文感知设备，并且最终确定异常事件 通过对正常情况和异常情况使用组合异常概率分布模型发生在监视区域。 （附图标记）（20）声音传感器; （30）图像传感器; （40）输入装置; （50）声音处理器; （60）图像处理器; （70）上下文感知设备; （80）输出装置
-
公开(公告)号：KR1020140015893A
公开(公告)日：2014-02-07
申请号：KR1020120081970
申请日：2012-07-26
Applicant: 한화테크윈 주식회사 , 고려대학교 산학협력단
CPC classification number: G01S3/8027 , G01H11/00 , G01S3/12 , G01S3/786 , G01S3/80 , G01S3/8086
Abstract: The present invention relates to a device and a method for estimating a sound source location and more specifically, to a device and a method for improving an estimation speed of a sound source location by reducing an SRP-PHAT calculation quantity for sound source location estimation. The method for estimating a sound source location comprises the steps of generating a histogram based on a plurality of interested abnormal sound frames to establish a database; fast Fourier transforming a sound signal inputted by a multichannel; selecting the certain number of frequency index from the fast Fourier transformed signal by using, as a control signal, the frequency index that meets the predetermined conditions in the histogram stored in the database; and calculating SRP-PHAT for a signal corresponding to the selected frequency index. The present invention is able to reduce an SRP-PHAT calculation quantity for sound source location estimation based on the database in which the histogram generated from the interested abnormal sound signal is stored, thereby improving an estimation speed of the sound source location. [Reference numerals] (AA) Start; (BB) End; (S10) Generate a histogram based on a plurality of interested abnormal sound frames to establish a database; (S20) Fast Fourier transform a sound signal inputted by a multichannel; (S30) Select the certain number of frequency index from the fast Fourier transformed signal by using, as a control signal, the frequency index that meets the predetermined conditions in the histogram stored in the database; (S40) Calculate SRP-PHAT for a signal corresponding to the selected frequency index; (S50) Accumulating the calculated SRP-PHAT for each certain time to estimate, to a sound source location, the direction where the maximum SRP-PHAT is accumulated the most
Abstract translation: 本发明涉及一种用于估计声源位置的装置和方法，更具体地，涉及一种通过减少用于声源位置估计的SRP-PHAT计算量来提高声源位置的估计速度的装置和方法。 用于估计声源位置的方法包括以下步骤：基于多个感兴趣的异常声音帧来生成直方图以建立数据库; 快速傅里叶变换由多声道输入的声音信号; 通过使用满足在数据库中存储的直方图中的预定条件的频率指标作为控制信号，从快速傅里叶变换信号中选择一定数量的频率索引; 并计算与所选频率指数对应的信号的SRP-PHAT。 本发明能够基于从感兴趣的异常声音信号生成的直方图被存储的数据库来减少用于声源位置估计的SRP-PHAT计算量，从而提高声源位置的估计速度。 （附图标记）（AA）开始; （BB）结束; （S10）基于多个感兴趣的异常声音帧生成直方图以建立数据库; （S20）快速傅里叶变换由多声道输入的声音信号; （S30）通过使用满足存储在数据库中的直方图中的预定条件的频率指标作为控制信号，从快速傅里叶变换信号中选择一定数量的频率索引; （S40）计算与选定频率指标对应的信号的SRP-PHAT; （S50）累积计算出的每个特定时间的SRP-PHAT以将声音源位置估计最大SRP-PHAT累加的方向最多
-
Patent Agency Ranking
公开(公告)号：KR101357381B1
公开(公告)日：2014-02-05
申请号：KR1020120050478
申请日：2012-05-11
Applicant: 고려대학교 산학협력단
IPC: G10L21/02
Abstract: 음성 인식을 위한 음성 인식 대상 구간을 검출하는 신호 처리 장치 및 방법이 제공된다. 상기 신호 처리 장치는, 입력 신호로부터 노이즈를 제거하여 제1 음성 인식 대상 구간을 추출하는 음성 추출부, 상기 입력 신호로부터 상기 신호 처리 장치에 연관된 반향이 포함된 구간을 제거하여 제2 음성 인식 대상 구간을 추출하는 반향 제거부, 및 상기 제1 음성 인식 대상 구간과 상기 제2 음성 인식 대상 구간의 중첩 구간을 상기 음성 인식을 위한 음성 인식 대상 구간으로 결정하는 연산부를 포함한다.
-
公开(公告)号：KR101339578B1
公开(公告)日：2013-12-10
申请号：KR1020120050706
申请日：2012-05-14
Applicant: 고려대학교 산학협력단
IPC: G10L15/00
Abstract: 본 발명은 잡음과 음성이 혼합된 환경에서 주파수 변환 및 원점 대칭 필터를 이용하여 음성 구간을 검출하는 음성 인식 방법 및 장치에 관한 것이다.
본 발명의 일실시예에 의한 음성 인식 방법은 잡음과 음성 신호가 혼합된 시간 영역의 오디오 입력 신호를 수신하는 단계; 상기 시간 영역의 오디오 입력 신호를 주파수 영역의 신호로 변환하는 단계; 음성 부재 확률에 대한 음성 존재 확률의 비를 이용하여 상기 오디오 입력 신호의 특징값을 추출하는 단계-상기 음성 존재 확률 및 음성 부재 확률은 상기 변환된 주파수 영역의 신호를 이용하여 획득됨-; 상기 추출된 특징값에 원점 대칭 필터를 적용하여 음성 구간을 검출하는 단계를 포함할 수 있다.-
公开(公告)号：KR1020130127077A
公开(公告)日：2013-11-22
申请号：KR1020120050706
申请日：2012-05-14
Applicant: 고려대학교 산학협력단
IPC: G10L15/00
Abstract: The present invention relates to a speech recognizing method and a device thereof detecting a speech section by using frequency conversion and an origin symmetric filter in an environment where noise and speech are mixed. The speech recognizing method comprises a step of receiving an audio input signal in a time domain where noise and a speech signal are mixed; a step of converting the audio input signal in the time domain into a signal in a frequency domain; a step of extracting a feature value of the audio input signal by using the ratio of speech absence probability to speech presence probability, obtaining the speech presence probability and the speech absence probability by using the converted signal in the frequency domain,; and a step of detecting a speech section by applying an origin symmetric filter to the extracted feature value. [Reference numerals] (110) Receiving unit;(120) Conversion unit;(130) Feature value extracting unit;(140) Voice section detecting unit;(150) Control unit
Abstract translation: 本发明涉及语音识别方法及其在混合噪声和语音的环境中通过使用频率转换和原始对称滤波器来检测语音部分的装置。 语音识别方法包括：在噪声和语音信号混合的时域中接收音频输入信号的步骤; 将时域中的音频输入信号转换为频域中的信号的步骤; 通过使用语音存在概率与语音存在概率的比率来提取音频输入信号的特征值，通过使用频域中的转换信号获得语音存在概率和语音不存在概率的步骤; 以及通过对所提取的特征值应用原始对称滤波器来检测语音部分的步骤。 （110）接收单元;（120）转换单元;（130）特征值提取单元;（140）语音区间检测单元;（150）控制单元
-
公开(公告)号：KR1020130046779A
公开(公告)日：2013-05-08
申请号：KR1020110111348
申请日：2011-10-28
Applicant: 고려대학교 산학협력단
IPC: G01S3/80
Abstract: PURPOSE: A sound source direction estimating device and a method thereof are provided to estimate the direction of sound transferred from one sound source in a vehicle which is equipped with at least three acoustic sensors. CONSTITUTION: Three acoustic sensors(310a,310b,310c) receive sound from the outside and output electric signals corresponding to the received sound. Based on the electric signals outputted from the three acoustic sensors, a TDOA estimating unit(320) estimates the TDOA for each pair of the acoustic sensor. A sound source direction estimating unit(330) estimates the direction of a desired sound source by performing a minimum filtering for the correlation values corresponding to a pair of all acoustic sensors estimated. A display unit(340) indicates the direction of the sound source estimated by the sound source direction estimating unit. [Reference numerals] (310a) Acoustic sensor 1; (310b) Acoustic sensor 2; (310c) Acoustic sensor 3; (320) TDOA estimating unit; (330) Sound source direction estimating unit; (340) Display unit
Abstract translation: 目的：提供一种声源方向估计装置及其方法，用于估计装备有至少三个声传感器的车辆中从一个声源转移的声音的方向。 构成：三声传感器（310a，310b，310c）从外部接收声音并输出与接收到的声音对应的电信号。 基于从三个声学传感器输出的电信号，TDOA估计单元（320）估计每对声学传感器的TDOA。 声源方向估计单元（330）通过对与估计的一对全部声学传感器对应的相关值执行最小滤波来估计期望声源的方向。 显示单元（340）指示由声源方向估计单元估计的声源的方向。 （附图标记）（310a）声传感器1; （310b）声传感器2; （310c）声传感器3; （320）TDOA估计单元; （330）声源方向估计单元; （340）显示单元
-
公开(公告)号：KR1020120020008A
公开(公告)日：2012-03-07
申请号：KR1020100083702
申请日：2010-08-27
Applicant: 한화테크윈 주식회사 , 고려대학교 산학협력단
CPC classification number: G06T3/4053
Abstract: PURPOSE: A very high resolution restoring method and an illegal parking checking system using the same are provided to combine pixel information of a low resolution image inputted at a current time with pixel information of a previously estimated very high resolution image, thereby restoring a high resolution image at the current time. CONSTITUTION: An illegal parking checking system receives a low resolution input image(S110). The system generates a high resolution estimated image(S120). The system compensates for a restoration error due to wrong image alignment wherein the restoration error is between the input image and the estimated error(S130). When a scene is suddenly changed, the system initializes a previously estimated high resolution target image(S140).
Abstract translation: 目的：提供一种非常高分辨率的恢复方法和使用该方法的非法停车检查系统，以将当前时间输入的低分辨率图像的像素信息与先前估计的非常高分辨率图像的像素信息相结合，从而恢复高分辨率 图像在当前时间。 规定：非法停车检查系统接收低分辨率输入图像（S110）。 该系统产生高分辨率估计图像（S120）。 该系统补偿由于错误的图像对准而导致的恢复误差，其中恢复误差在输入图像与估计误差之间（S130）。 当场景突然改变时，系统初始化先前估计的高分辨率目标图像（S140）。
-
-
-
-
-
-
-
-
-
Search Results
97
records
(took 0.032 seconds)
