公开(公告)号：US5727124A

公开(公告)日：1998-03-10

申请号：US263284

申请日：1994-06-21

申请人： Chin-Hui Lee ,  Ananth Sankar

发明人： Chin-Hui Lee ,  Ananth Sankar

IPC分类号： G06F15/18 ,  G06N3/00 ,  G10L15/02 ,  G10L15/06 ,  G10L15/08 ,  G10L15/14 ,  G10L15/16 ,  G10L15/20 ,  G10L15/22 ,  G10L19/00 ,  G10L21/02 ,  G10L5/00

CPC分类号： G10L15/065 ,  G10L15/14 ,  G10L25/27

摘要： Disclosed is a method for drastically reducing the average error rate for signals under mismatched conditions. The method takes a signal (e.g., speech signal) and a set of stored representations (e.g., stored representations of keywords) and performs at least one transformation that results in the signal more closely emulating the stored representations. This is accomplished by using one of three techniques. First, one may transform the signal so that the signal may be better approximated by (e.g., is closer to) one of the stored representations. Second, one may transform the set of stored representations so that one of the stored representations better approximates the signal. Third, one may transform both the signal and the set of stored representations.

摘要翻译： 公开了一种用于在不匹配条件下显着降低信号的平均误码率的方法。 该方法获取信号（例如，语音信号）和一组存储的表示（例如，存储的关键字的表示），并执行至少一个导致信号更接近地仿真所存储的表示的变换。 这通过使用三种技术之一来实现。 首先，可以对信号进行变换，使得信号可以通过（例如，更靠近）存储的表示之一更好地近似。 第二，可以转换所存储的表示集合，使得所存储的表示中的一个更接近于该信号。 第三，可以转换信号和存储的表示集合。

公开(公告)号：US5680507A

公开(公告)日：1997-10-21

申请号：US564611

申请日：1995-11-29

申请人： Juin-Hwey Chen

发明人： Juin-Hwey Chen

IPC分类号： H03M1/12 ,  G10L11/02 ,  G10L11/04 ,  G10L11/06 ,  G10L19/00 ,  G10L19/08 ,  G10L19/12 ,  G10L19/14 ,  H03M7/00 ,  H03M7/30 ,  G10L5/00

CPC分类号： G10L19/18 ,  G10L19/08 ,  G10L19/12 ,  G10L19/26 ,  G10L2025/786 ,  G10L2025/906 ,  G10L25/06 ,  G10L25/93

摘要： Codebook vectors may be considered critical if they give poor energy approximations and exhibit a particular shape with smaller components near the beginning and larger components toward the end of the vector. Standard deviation may be used to identify critical codevectors based on energy approximation error measured in decibels. A low-bit rate (typically 8 kbit/s or less), low-delay digital coder and decoder based on Code Excited Linear Prediction for speech and similar signals features backward adaptive adjustment for codebook gain and short-term synthesis filter parameters and forward adaptive adjustment of long-term (pitch) synthesis filter parameters. In addition, the coder makes use of an excitation codebook and the coding is based on a set of codebook vector energies for a set of codebook vectors in the codebook. The codebook energies are calculated by identifying a set of approximations for the non-critical codebook vector energies. This achieves a significant reduction in processing time in comparison with prior art techniques.

摘要翻译： 如果代码本向量给出较差的能量近似值并且表现出特定形状，并且在开头附近具有较小的分量并且朝向向量的末端具有较大的分量，则这些矢量可能被认为是关键 标准偏差可用于基于以分贝为单位测量的能量近似误差来识别关键代码矢量。 基于用于语音和类似信号的码激励线性预测的低比特率（通常为8kbit / s或更小），低延迟数字编码器和解码器特征在于对于码本增益和短期合成滤波器参数以及前向自适应 调整长期（间距）合成滤波参数。 此外，编码器使用激励码本，并且编码基于码本中的一组码本矢量的一组码本矢量能量。 通过识别非关键码本矢量能量的一组近似来计算码本能量。 与现有技术相比，这实现了处理时间的显着减少。

公开(公告)号：US5680505A

公开(公告)日：1997-10-21

申请号：US715750

申请日：1996-09-19

申请人： Kit-fun Ho

发明人： Kit-fun Ho

IPC分类号： G10L15/24 ,  G10L15/26 ,  G10L5/00

CPC分类号： G10L15/24 ,  G10L15/26

摘要： A plurality of transducers are positioned in front of a speaker's mouth for detecting and responding to air flow patterns in space and time. Specific examples for the system and method for speech analysis and recognition by the detection of air flow pattern in the proximity of the mouth in space and time during an utterance are provided.

摘要翻译： 多个换能器位于扬声器嘴的前面，用于在空间和时间中检测和响应空气流动模式。 提供了用于语音分析和识别的系统和方法的具体示例，其通过在发音期间在空间和时间中在口附近检测气流模式。

公开(公告)号：US5644677A

公开(公告)日：1997-07-01

申请号：US120266

申请日：1993-09-13

申请人： Sangil Park ,  Dion M. Funderburk

发明人： Sangil Park ,  Dion M. Funderburk

IPC分类号： G10H1/20 ,  G10H1/36 ,  G10K15/04 ,  G10L21/04 ,  G10L5/00

CPC分类号： G10L21/01 ,  G10H1/20 ,  G10H1/366 ,  G10H2250/115 ,  G10H2250/121 ,  G10H2250/621

摘要： A signal processing system (50) performs real-time pitch shifting for applications such as karaoke, tapeless answering machines, and the like while minimizing distortion. A digital input signal is sampled and stored at successive locations in a variable-size buffer (62) at an input sample rate. Data from the variable-size buffer (62) is interpolated according to a pitch-shifting ratio. An adaptive pitch estimator (61) continually estimates the fundamental frequency of the digital input signal, and the signal processing system (50) adjusts the buffer size of the variable-size buffer (62) in response thereto. The signal processing system (50) changes the buffer size to store the digital input signal for an integral number of periods of the estimated fundamental frequency.

摘要翻译： 信号处理系统（50）在使失真最小化的同时对诸如卡拉OK，无电话应答机等的应用执行实时音调移位。 数字输入信号被采样并以输入采样率存储在可变大小缓冲器（62）中的连续位置处。 来自可变尺寸缓冲器（62）的数据根据​​音调偏移比进行内插。 自适应音调估计器（61）连续地估计数字输入信号的基频，并且信号处理系统（50）响应于此调整可变大小缓冲器（62）的缓冲器大小。 信号处理系统（50）改变缓冲器大小以存储所估计的基频的整数个周期的数字输入信号。

公开(公告)号：US5640488A

公开(公告)日：1997-06-17

申请号：US435882

申请日：1995-05-05

申请人： Jean-claude Junqua ,  Craig Demel

发明人： Jean-claude Junqua ,  Craig Demel

IPC分类号： A61N5/04 ,  G06F17/30 ,  G10L15/06 ,  G10L5/00

CPC分类号： G06F17/30952 ,  A61N5/04 ,  G10L2015/0631

摘要： The dictionary is broken into clusters by first grouping the dictionary according to a rule based procedure whereby the dictionary is sorted by word length and alphabetically. After sorting, a plurality of first cluster centers is generated by selecting the dictionary entries that differ from neighboring entries by the first letter. Each of the dictionary entries is then assigned to the closest one of the first cluster centers using a dynamic time warping procedure. These newly formed clusters are then each analyzed to find the true cluster center and the dictionary entries are then each assigned to the closest true cluster center. The clusters, so formed, may then be rapidly searched to locate any dictionary entry. The search is quite efficient because only the closest cluster to the desired dictionary entry needs to be searched.

摘要翻译： 通过首先根据基于规则的过程对字典进行分组，将词典分解成集群，由此字典按字长和字母排序。 在排序之后，通过选择与第一个字母相邻的条目不同的字典条目来生成多个第一集群中心。 然后使用动态时间扭曲过程将每个字典条目分配给最接近的一个第一集群中心。 然后对这些新形成的集群进行分析，以找到真正的集群中心，然后将字典条目分配给最接近的真实集群中心。 然后可以快速搜索如此形成的群集以定位任何字典条目。 搜索是非常有效的，因为只需要搜索到所需字典条目的最近的集群。

公开(公告)号：US5638486A

公开(公告)日：1997-06-10

申请号：US329394

申请日：1994-10-26

申请人： Shay-Ping T. Wang ,  Michael K. Lindsey

发明人： Shay-Ping T. Wang ,  Michael K. Lindsey

IPC分类号： G10L15/02 ,  G10L15/10 ,  G10L5/00

CPC分类号： G10L15/10 ,  G10L15/02

摘要： In a speech-recognition system having a plurality of classifiers, a voting window includes a sequence of outputs from each of the classifiers. For each classifier, a voting sum is generated corresponding to the voting window. A spoken sound is identified by determining which classifier corresponds to the greatest voting sum.

摘要翻译： 在具有多个分类器的语音识别系统中，投票窗口包括来自每个分类器的输出序列。 对于每个分类器，对应于投票窗口生成投票总和。 通过确定哪个分类器对应于最大投票总和来识别口语声音。

公开(公告)号：US5634084A

公开(公告)日：1997-05-27

申请号：US376732

申请日：1995-01-20

申请人： Bathsheba J. Malsheen ,  Gabriel F. Groner ,  Sandra F. Disner

发明人： Bathsheba J. Malsheen ,  Gabriel F. Groner ,  Sandra F. Disner

IPC分类号： G06F17/27 ,  G06F17/28 ,  G10L13/08 ,  G10L5/00

CPC分类号： G06F17/273 ,  G06F17/2755 ,  G06F17/278 ,  G06F17/2881 ,  G06F17/289 ,  G10L13/08

摘要： An improved text-to-speech synthesizer that employs a text to speech converter, a text reader control procedure, a classifier procedure, an abbreviation expansion procedure, and an acronym/initialism expanding procedure is herein described. A classifier procedure is used to classify generate classification values for each word in the text message with regard to syntax, punctuation and membership in predefined classes of words, the predefined classes of words including number, measurement units, geographic designations, and date/time values. An abbreviation expansion procedure evaluates, based on the classification values for words neighboring the identified words, which, if any, of the potential expansion values is applicable, and substitutes the potential expansion for the identified abbreviation word when evaluation yields a success value. An acronym/initialism expanding procedure identifies words in the text message that are acronyms and initialisms, parses pronounceable syllables within the identified words and generates a substitute string that can consist of any combination of letters, numbers, pronounceable syllables or multiple letter identifiers.

摘要翻译： 本文描述了采用文本到语音转换器的改进的文本到语音合成器，文本读取器控制过程，分类器过程，缩写扩展过程和首字母缩略词/初始扩展过程。 分类器过程用于对文本消息中关于语言，标点符号和预定义词类中的成员资格的每个单词的生成分类值进行分类，包括数字，测量单位，地理名称和日期/时间值的预定义词类 。 缩写扩展程序根据所识别的单词相邻的单词的分类值进行评估，如果有的话，潜在的扩展值是可适用的，并且当评估产生成功值时，替代所识别的缩写词的潜在扩展。 首字母缩略词/初始化扩展过程识别文本消息中的首字母缩略词和初始化词，在所标识的单词中解析可发音的音节，并生成可由字母，数字，可发音音节或多个字母标识符的任意组合组成的替代字符串。

公开(公告)号：US5634083A

公开(公告)日：1997-05-27

申请号：US203105

申请日：1994-02-28

申请人： Martin Oerder

发明人： Martin Oerder

IPC分类号： G10L15/22 ,  G10L15/04 ,  G10L15/08 ,  G10L15/12 ,  G10L15/18 ,  G10L15/28 ,  G10L9/00 ,  G10L5/00

CPC分类号： G10L15/083

摘要： The procedure for the recognition of a speech signal until output of the recognized word sequence or the recognized sentence is split in accordance with the invention in such a manner that first only word hypotheses are separately generated for different starting instants and that from these word hypotheses preliminary word strings are formed in conformity with a word graph, the word graph thus arising being continuously optimized by erasure of parts of word strings. Parts of word strings having the same beginning and end points are compared with one another and the scores of words having concurrent end points are compared with a threshold value. Further steps for optimization of the word graph are also shown. For output disclosed a particularly effective post-editing operation where for each incorrect word all further words having the same beginning are output, enabling fast selection of the correct word from all said further words, by the operator.

摘要翻译： 用于识别语音信号的过程直到按照本发明的方式将所识别的字序列或识别的句子的输出分开，以便对于不同的起始时刻分别产生第一个单词假设，并且从这些单词假设初步 字符串形成为符合字图，因此通过擦除字串的部分不断优化出现的字图。 将具有相同起始点和终点的字串的部分彼此进行比较，并将具有并发终点的单词的分数与阈值进行比较。 还显示了用于优化字图的进一步步骤。 对于输出，公开了一种特别有效的后编辑操作，其中对于每个不正确的单词，输出具有相同开始的所有进一步的单词，从而能够由操作者快速地从所有所述另外的单词中选择正确的单词。

公开(公告)号：US5583784A

公开(公告)日：1996-12-10

申请号：US241851

申请日：1994-05-12

申请人： Rolf Kapust ,  Dieter Seltzer

发明人： Rolf Kapust ,  Dieter Seltzer

IPC分类号： G10L25/27 ,  G10L25/48 ,  H04B1/66 ,  G10L5/00 ,  G10L7/02 ,  G10L9/00

CPC分类号： H04B1/665 ,  G10L25/48 ,  G10L25/27

摘要： A frequency analysis method comprises using a window function to evaluate aemporal input signal present in the form of discrete sampled values. The windowed input signal is subsequently subjected to Fourier transformation for the purpose of generating a set of coefficients. In order to develop such a method so that the characteristics of the human ear are simulated not only with respect to the spectral projection in the frequency range, but also with respect to the resolution in the temporal range, a set of different window functions is used to evaluate a block of the input signal in order to generate a set of blocks, weighted with the respective window functions, of sampled values whose Fourier transforms have different bandwidths, before each of the simultaneously generated blocks of sampled values is subjected to a dedicated Fourier transformation in such a way that for each window function at least respectively one coefficient is calculated which is assigned the bandwidth of the Fourier transforms of this window function, and that the coefficients are chosen such that the frequency bands assigned to them essentially adjoin one another.

摘要翻译： 频率分析方法包括使用窗口函数来评估以离散采样值的形式存在的时间输入信号。 随后，窗口化的输入信号经历傅里叶变换，以产生一组系数。 为了开发这样的方法，使得不仅相对于频率范围内的光谱投影而且对于时间范围内的分辨率来模拟人耳的特性，所以使用一组不同的窗函数 为了评估输入信号的块，以便在每个采样值的每个同时产生的采样值块经受专用傅里叶变换之前，生成一组利用各个窗口函数加权的块，其中傅立叶变换具有不同带宽的采样值 以这样的方式进行变换，使得对于每个窗口函数，至少分别计算一个被分配了该窗口函数的傅立叶变换的带宽的系数，并且选择这些系数使得分配给它们的频带基本上彼此相邻。

公开(公告)号：US5465378A

公开(公告)日：1995-11-07

申请号：US135167

申请日：1993-10-12

申请人： David T. Duensing ,  Dayle E. Phillips ,  Dennis C. Simmons

发明人： David T. Duensing ,  Dayle E. Phillips ,  Dennis C. Simmons

IPC分类号： G06F3/16 ,  G10L15/22 ,  G06F15/00 ,  G10L5/00

CPC分类号： G06F3/16 ,  G10L15/22

摘要： A preferred report generating system includes a computer (12) responsive to user-spoken inputs for selecting previously defined report material including text and graphics stored in memory respectively corresponding to the inputs, for activating other user inputs, and implementing corresponding computer commands. After receipt of preferred user-spoken inputs entered by way of a microphone (16) representing information needed for generating a report, the system compiles the report material corresponding to the user-selected inputs for generating the report.

摘要翻译： 优选的报告生成系统包括响应于用户口头输入的计算机（12），用于选择先前定义的报告材料，包括分别对应于输入的存储在存储器中的文本和图形，以激活其他用户输入，以及实现相应的计算机命令。 在收到通过表示生成报告所需的信息的麦克风（16）输入的优选用户口头输入之后，系统编译与用户选择的输入相对应的用于生成报告的报告材料。