公开(公告)号：US4361875A

公开(公告)日：1982-11-30

申请号：US162261

申请日：1980-06-23

申请人： David Hertz ,  Robert P. Kurshan ,  David Malah

发明人： David Hertz ,  Robert P. Kurshan ,  David Malah

IPC分类号： G01R23/15 ,  H04Q1/457 ,  G06F15/31

CPC分类号： G01R23/15 ,  H04Q1/457

摘要： A system (1000) for estimating the frequency of a tone input utilizes sample rate reduction in successive stages and processing by digital cyclotomic filters at each stage. The tone input (2001) is first transformed in network (1100) to yield two quadrature tones. Digitizer (1200) converts the two tones into data words.Buffer (1300) comprises two essentially identical storage arrangements wherein data words are stored and then supplied to succeeding stages. Frequency-shifting unit (1400) effects modulus-one multiplication by processing appropriately selected data words. Word pairs and frequency-shifted versions thereof are processed by cyclotomic filters (1500). Sequential decimation in this system effects a successively refined estimate to the tonal frequency. During each stage of decimation, the filters are configured to provide symmetric coverage of the subband containing the estimate. Configuration information is provided by decision unit (1600) via threshold comparison of the outputs from the filters and controller (1700) provides control information to the elements of the system. Rate reduction occurs on a 2:1 basis for each stage of decimation. The first frequency interval covered is one-fourth the initial sampling rate, and each stage of decimation causes a factor of two refinement in the estimate. The filters are structured as the equivalent of two pairs of first-order filters at each stage of decimation.

摘要翻译： 用于估计音调输入频率的系统（1000）在连续的阶段中使用采样率降低，并且在每个阶段利用数字循环滤波器进行处理。 音调输入（2001）首先在网络（1100）中变换以产生两个正交音调。 数字转换器（1200）将两个音调转换为数据字。 缓冲器（1300）包括两个基本上相同的存储装置，其中数据字被存储然后提供给后续的级。 频移单元（1400）通过处理适当选择的数据字来影响模数一乘法。 字对及其频移版本由循环滤波器（1500）处理。 该系统中的连续抽取对音调频率产生了连续精确的估计。 在抽取的每个阶段期间，滤波器被配置为提供包含估计的子带的对称覆盖。 配置信息由决策单元（1600）通过来自滤波器的输出的阈值比较提供，并且控制器（1700）将控制信息提供给系统的元件。 降频发生在每个抽取阶段的2：1基础上。 覆盖的第一个频率间隔是初始采样率的四分之一，并且每个抽取阶段在估计中导致两个细化因子。 滤波器在抽取的每个阶段被构造为等价于两对一阶滤波器。

公开(公告)号：US4348735A

公开(公告)日：1982-09-07

申请号：US162262

申请日：1980-06-23

申请人： David Hertz ,  Robert P. Kurshan ,  David Malah

发明人： David Hertz ,  Robert P. Kurshan ,  David Malah

IPC分类号： G01R23/167 ,  H03H17/02 ,  G06F15/31

CPC分类号： H03H17/02 ,  G01R23/167

摘要： A system (1000) for estimating the frequency of a tone input utilizes sample rate reduction in successive stages and processing by digital cyclotomic filters at each stage. The tone input (2001) is first transformed in network (1100) to yield two quadrature tones. Digitizer (1200) converts the two tones into data words. Buffer (1300) comprises two essentially identical storage arrangments wherein data words are stored and then supplied to succeeding stages. Frequency-shifting unit (1400) effects modulus-one multiplication by processing appropriately selected data words. Word pairs and frequency-shifted versions thereof are processed by cyclotomic filters (1500). Sequential decimation in the system effects a successively refined estimate to the tonal frequency. During each stage of decimation, the filters are configured to provide symmetric coverage of the subband containing the estimate. Configuration information is provided by decision unit (1600) via threshold comparison of the outputs from the filters and controller (1700) provides control information to the elements of the system. Rate reduction occurs on a 4:1 basis for each stage of decimation. The first frequency interval covered is one-fourth the initial sampling rate, and each stage of decimation causes a factor of four refinement in the estimate. The filters are structured as the equivalent of four filter pairs at each stage of decimation; two of the four pairs are of first order, whereas the other two are of second order.

摘要翻译： 用于估计音调输入频率的系统（1000）在连续的阶段中使用采样率降低，并且在每个阶段利用数字循环滤波器进行处理。 音调输入（2001）首先在网络（1100）中变换以产生两个正交音调。 数字转换器（1200）将两个音调转换为数据字。 缓冲器（1300）包括两个基本上相同的存储布置，其中存储数据字，然后提供给后续级。 频移单元（1400）通过处理适当选择的数据字来影响模数一乘法。 字对及其频移版本由循环滤波器（1500）处理。 系统中的顺序抽取对音调频率产生了连续精确的估计。 在抽取的每个阶段期间，滤波器被配置为提供包含估计的子带的对称覆盖。 配置信息由决策单元（1600）通过来自滤波器的输出的阈值比较提供，并且控制器（1700）将控制信息提供给系统的元件。 降频发生在每个抽取阶段的4比1。 覆盖的第一个频率间隔是初始采样率的四分之一，并且每个抽取阶段在估计中导致四个细化因子。 滤波器在抽取的每个阶段被构造为等效于四个滤波器对; 四对中的两个是第一顺序，而另外两个是二阶。

公开(公告)号：US4333156A

公开(公告)日：1982-06-01

申请号：US162256

申请日：1980-06-23

申请人： Robert P. Kurshan ,  David Malah

发明人： Robert P. Kurshan ,  David Malah

IPC分类号： G06F17/10 ,  G06F15/31

CPC分类号： G06F17/10

摘要： A system (1000) for estimating the frequency of a tone input utilizes sample rate restriction in successive stages and processing by digital cyclotomic filters at each stage. The tone input (2001) is first transformed in network (1100) to yield two quadrature tones. Digitizer (1200) converts the two tones into data words. Buffer (1300) comprises two essentially identical storage arrangements wherein data words are stored and then supplied to succeeding stages. Frequency-shifting unit (1400) effects modulus-one multiplication by processing appropriately selected data words. Word pairs and frequency-shifted versions thereof are processed by cyclotomic filters (1500). Sequential decimation in the system effects a successively refined estimate to the tonal frequency. During each stage of decimation, the filters are configured to provide symmetric coverage of the subband containing the estimate. Configuration information is provided by decision unit (1600) via threshold comparison of the outputs from the filters and controller (1700) provides control information to the elements of the system. Rate reduction occurs on a 4:1 basis for each stage of decimation. The first frequency interval covered is one-fourth the initial sampling rate, and each stage of decimation causes a factor of four refinement in the estimate. The filters are structured as the equivalent of four filter pairs at each stage of decimation, two of the four pairs are of first order, whereas the other two are of second order.

摘要翻译： 用于估计音调输入频率的系统（1000）在连续的阶段中利用采样率限制，并在每个阶段利用数字循环滤波器进行处理。 音调输入（2001）首先在网络（1100）中变换以产生两个正交音调。 数字转换器（1200）将两个音调转换为数据字。 缓冲器（1300）包括两个基本上相同的存储装置，其中数据字被存储然后提供给后续的级。 频移单元（1400）通过处理适当选择的数据字来影响模数一乘法。 字对及其频移版本由循环滤波器（1500）处理。 系统中的顺序抽取对音调频率产生了连续精确的估计。 在抽取的每个阶段期间，滤波器被配置为提供包含估计的子带的对称覆盖。 配置信息由决策单元（1600）通过来自滤波器的输出的阈值比较提供，并且控制器（1700）将控制信息提供给系统的元件。 降频发生在每个抽取阶段的4比1。 覆盖的第一个频率间隔是初始采样率的四分之一，并且每个抽取阶段在估计中导致四个细化因子。 滤波器在抽取的每个阶段被构造为等效于四个滤波器对，四对中的两个是一阶的，而另外两个是二阶的。

公开(公告)号：US07613604B1

公开(公告)日：2009-11-03

申请号：US11691160

申请日：2007-03-26

申请人： David Malah ,  Richard Vandervoort Cox

发明人： David Malah ,  Richard Vandervoort Cox

IPC分类号： G10L21/00

CPC分类号： G10L21/038

摘要： A system and method are disclosed for extending the bandwidth of a narrowband signal such as a speech signal. The method applies a parametric approach to bandwidth extension but does not require training. The parametric representation relates to a discrete acoustic tube model (DATM). The method comprises computing narrowband linear predictive coefficients (LPCs) from a received narrowband speech signal, computing narrowband partial correlation coefficients (parcors) using recursion, computing Mnb area coefficients from the partial correlation coefficient, and extracting Mwb area coefficients using interpolation. Wideband parcors are computed from the Mwb area coefficients and wideband LPCs are computed from the wideband parcors. The method further comprises synthesizing a wideband signal using the wideband LPCs and a wideband excitation signal, highpass filtering the synthesized wideband signal to produce a highband signal, and combining the highband signal with the original narrowband signal to generate a wideband signal. In a preferred variation of the invention, the Mnb area coefficients are converted to log-area coefficients for the purpose of extracting, through shifted-interpolation, Mwb log-area coefficients. The Mwb log-area coefficients are then converted to Mwb area coefficients before generating the wideband parcors.

摘要翻译： 公开了用于扩展诸如语音信号的窄带信号的带宽的系统和方法。 该方法对带宽扩展采用参数化方法，但不需要培训。 参数表示涉及离散声管模型（DATM）。 该方法包括从接收的窄带语音信号中计算窄带线性预测系数（LPC），使用递归计算窄带部分相关系数（parcors），从部分相关系数计算Mnb面积系数，以及使用插值提取Mwb面积系数。 从Mwb区域系数计算宽带掩码，并从宽带掩码计算宽带LPC。 该方法还包括使用宽带LPC和宽带激励信号合成宽带信号，对合成的宽带信号进行高通滤波以产生高频带信号，以及将高频带信号与原始窄带信号组合以产生宽带信号。 在本发明的优选变型中，Mnb面积系数被转换为对数面积系数，以便通过移位插值提取Mwb对数面积系数。 然后在生成宽带掩码之前，将Mwb对数区域系数转换为Mwb区域系数。

公开(公告)号：US08595001B2

公开(公告)日：2013-11-26

申请号：US13290464

申请日：2011-11-07

申请人： David Malah ,  Richard Vandervoort Cox

发明人： David Malah ,  Richard Vandervoort Cox

IPC分类号： G10L19/00

CPC分类号： G10L21/038

摘要： A method applies a parametric approach to bandwidth extension but does not require training. The method computes narrowband linear predictive coefficients from a received narrowband speech signal, computes narrowband partial correlation coefficients using recursion, computes Mnb area coefficients from the partial correlation coefficient, and extracts Mwb area coefficients using interpolation. Wideband parcors are computed from the Mwb area coefficients and wideband LPCs are computed from the wideband parcors. The method further comprises synthesizing a wideband signal using the wideband LPCs and a wideband excitation signal, highpass filtering the synthesized wideband signal to produce a highband signal, and combining the highband signal with the original narrowband signal to generate a wideband signal.

摘要翻译： 一种方法将参数化方法应用于带宽扩展，但不需要培训。 该方法从接收的窄带语音信号计算窄带线性预测系数，使用递归计算窄带部分相关系数，从部分相关系数计算Mnb面积系数，并使用插值提取Mwb面积系数。 从Mwb区域系数计算宽带掩码，并从宽带掩码计算宽带LPC。 该方法还包括使用宽带LPC和宽带激励信号合成宽带信号，对合成的宽带信号进行高通滤波以产生高频带信号，以及将高频带信号与原始窄带信号组合以产生宽带信号。

公开(公告)号：US20100042408A1

公开(公告)日：2010-02-18

申请号：US12582034

申请日：2009-10-20

申请人： David Malah ,  Richard Vandervoort Cox

发明人： David Malah ,  Richard Vandervoort Cox

IPC分类号： G10L21/00

CPC分类号： G10L21/038

摘要： A system and method are disclosed for extending the bandwidth of a narrowband signal such as a speech signal. The method applies a parametric approach to bandwidth extension but does not require training. The parametric representation relates to a discrete acoustic tube model (DATM). The method comprises computing narrowband linear predictive coefficients (LPCs) from a received narrowband speech signal, computing narrowband partial correlation coefficients (parcors) using recursion, computing Mnb area coefficients from the partial correlation coefficient, and extracting Mwb area coefficients using interpolation. Wideband parcors are computed from the Mwb area coefficients and wideband LPCs are computed from the wideband parcors. The method further comprises synthesizing a wideband signal using the wideband LPCs and a wideband excitation signal, highpass filtering the synthesized wideband signal to produce a highband signal, and combining the highband signal with the original narrowband signal to generate a wideband signal. In a preferred variation of the invention, the Mnb area coefficients are converted to log-area coefficients for the purpose of extracting, through shifted-interpolation, Mwb log-area coefficients. The Mwb log-area coefficients are then converted to Mwb area coefficients before generating the wideband parcors.

摘要翻译： 公开了用于扩展诸如语音信号的窄带信号的带宽的系统和方法。 该方法对带宽扩展采用参数化方法，但不需要培训。 参数表示涉及离散声管模型（DATM）。 该方法包括从接收的窄带语音信号中计算窄带线性预测系数（LPC），使用递归计算窄带部分相关系数（parcors），从部分相关系数计算Mnb面积系数，以及使用插值提取Mwb面积系数。 从Mwb区域系数计算宽带掩码，并从宽带掩码计算宽带LPC。 该方法还包括使用宽带LPC和宽带激励信号合成宽带信号，对合成的宽带信号进行高通滤波以产生高频带信号，以及将高频带信号与原始窄带信号组合以产生宽带信号。 在本发明的优选变型中，Mnb面积系数被转换为对数面积系数，以便通过移位插值提取Mwb对数面积系数。 然后在生成宽带掩码之前，将Mwb对数区域系数转换为Mwb区域系数。

公开(公告)号：US20120116769A1

公开(公告)日：2012-05-10

申请号：US13290464

申请日：2011-11-07

申请人： David Malah ,  Richard Vandervoort Cox

发明人： David Malah ,  Richard Vandervoort Cox

IPC分类号： G10L13/00

CPC分类号： G10L21/038

摘要： A method applies a parametric approach to bandwidth extension but does not require training. The method computes narrowband linear predictive coefficients from a received narrowband speech signal, computes narrowband partial correlation coefficients using recursion, computes Mnb area coefficients from the partial correlation coefficient, and extracts Mwb area coefficients using interpolation. Wideband parcors are computed from the Mwb area coefficients and wideband LPCs are computed from the wideband parcors. The method further comprises synthesizing a wideband signal using the wideband LPCs and a wideband excitation signal, highpass filtering the synthesized wideband signal to produce a highband signal, and combining the highband signal with the original narrowband signal to generate a wideband signal.

摘要翻译： 一种方法将参数化方法应用于带宽扩展，但不需要培训。 该方法从接收的窄带语音信号计算窄带线性预测系数，使用递归计算窄带部分相关系数，从部分相关系数计算Mnb面积系数，并使用插值提取Mwb面积系数。 从Mwb区域系数计算宽带掩码，并从宽带掩码计算宽带LPC。 该方法还包括使用宽带LPC和宽带激励信号合成宽带信号，对合成的宽带信号进行高通滤波以产生高频带信号，以及将高频带信号与原始窄带信号组合以产生宽带信号。

公开(公告)号：US08069038B2

公开(公告)日：2011-11-29

申请号：US12582034

申请日：2009-10-20

申请人： David Malah ,  Richard Vandervoort Cox

发明人： David Malah ,  Richard Vandervoort Cox

IPC分类号： G10L21/00

CPC分类号： G10L21/038

摘要： A system and method are disclosed for extending the bandwidth of a narrowband signal such as a speech signal. The method applies a parametric approach to bandwidth extension but does not require training. The parametric representation relates to a discrete acoustic tube model (DATM). The method comprises computing narrowband linear predictive coefficients (LPCs) from a received narrowband speech signal, computing narrowband partial correlation coefficients (parcors) using recursion, computing Mnb area coefficients from the partial correlation coefficient, and extracting Mwb area coefficients using interpolation. Wideband parcors are computed from the Mwb area coefficients and wideband LPCs are computed from the wideband parcors. The method further comprises synthesizing a wideband signal using the wideband LPCs and a wideband excitation signal, highpass filtering the synthesized wideband signal to produce a highband signal, and combining the highband signal with the original narrowband signal to generate a wideband signal. In a preferred variation of the invention, the Mnb area coefficients are converted to log-area coefficients for the purpose of extracting, through shifted-interpolation, Mwb log-area coefficients. The Mwb log-area coefficients are then converted to Mwb area coefficients before generating the wideband parcors.

摘要翻译： 公开了用于扩展诸如语音信号的窄带信号的带宽的系统和方法。 该方法对带宽扩展采用参数化方法，但不需要培训。 参数表示涉及离散声管模型（DATM）。 该方法包括从接收的窄带语音信号中计算窄带线性预测系数（LPC），使用递归计算窄带部分相关系数（parcors），从部分相关系数计算Mnb面积系数，以及使用插值提取Mwb面积系数。 从Mwb区域系数计算宽带掩码，并从宽带掩码计算宽带LPC。 该方法还包括使用宽带LPC和宽带激励信号合成宽带信号，对合成的宽带信号进行高通滤波以产生高频带信号，以及将高频带信号与原始窄带信号组合以产生宽带信号。 在本发明的优选变型中，Mnb面积系数被转换为对数面积系数，以便通过移位插值提取Mwb对数面积系数。 然后在生成宽带掩码之前，将Mwb对数区域系数转换为Mwb区域系数。

公开(公告)号：US20050187759A1

公开(公告)日：2005-08-25

申请号：US11113463

申请日：2005-04-25

申请人： David Malah ,  Richard Cox

发明人： David Malah ,  Richard Cox

IPC分类号： G10L13/04 ,  G10L19/02 ,  G10L19/06 ,  G10L21/00 ,  G10L21/02

CPC分类号： G10L21/038

摘要： A system, computer-readable medium and generated signal are disclosed for extending the bandwidth of a first signal (i.e., a narrowband signal) such as a speech signal. The system produces a second signal from a first signal by computing first area coefficients from a first signal, generating second area coefficients from the first area coefficients and generating a second signal using the second area coefficients. The first signal may be a narrowband signal and second signal may be a wideband signal. The first area coefficients may be narrowband coefficients and the second area coefficients may be wideband area coefficients.

摘要翻译： 公开了一种用于扩展诸如语音信号的第一信号（即，窄带信号）的带宽的系统，计算机可读介质和生成的信号。 该系统通过从第一信号计算第一区域系数产生来自第一信号的第二信号，从第一区域系数产生第二区域系数，并使用第二区域系数产生第二信号。 第一信号可以是窄带信号，第二信号可以是宽带信号。 第一区域系数可以是窄带系数，第二区域系数可以是宽带面积系数。

公开(公告)号：US5991718A

公开(公告)日：1999-11-23

申请号：US31726

申请日：1998-02-27

申请人： David Malah

发明人： David Malah

IPC分类号： G10L25/78 ,  G10L9/00

CPC分类号： G10L25/78 ,  G10L2025/786

摘要： The system and method of the invention relates to voice detection technology for determining instants of time at which a snapshot of noise characteristics results in improved adaptation of noise floors used in voice detection. The approach is based on the "lower envelope" of the smoothed input signal power. Incorporation of this approach in a simple time domain VAD (Voice Activity Detector) results in an effective low-complexity system which, on the basis of simulations, gives good performance down to SNR values of about 0 dB. In the invention the lower envelope also provides the updated value of the noise threshold during the presence of speech. The invention can also be embedded in other, more complex (e.g., frequency domain) VADs at low computational cost.

摘要翻译： 本发明的系统和方法涉及用于确定时间的瞬间的语音检测技术，其中噪声特征的快照导致在语音检测中使用的噪声底层的改进的适应。 该方法基于平滑的输入信号功率的“下限”。 将这种方法结合在简单的时域VAD（语音活动检测器）中产生了一种有效的低复杂度系统，其在模拟的基础上提供了低于约0dB的SNR值的良好性能。 在本发明中，下部信封还在语音存在期间提供噪声阈值的更新值。 本发明也可以以低的计算成本嵌入在其他更复杂（例如，频域）VAD中。