Abstract: Systems and methods improving the performance of statistical model-based single-channel speech enhancement systems using a deep neural network (DNN) are disclosed. Embodiments include a DNN-trained system to predict speech presence in the input signal, and this information can be used to create frameworks for tracking noise and conducting a priori signal to-noise ratio estimation. Example frameworks provide increased flexibility for various aspects of system design, such as gain estimation. Examples include training a DNN to detect speech in the presence of both noise and reverberation, enabling joint suppression of additive noise and reverberation. Example frameworks provide significant improvements in objective speech quality metrics relative to baseline systems.

Abstract: Systems and methods improving the performance of statistical model-based single-channel speech enhancement systems using a deep neural network (DNN) are disclosed. Embodiments include a DNN-trained system to predict speech presence in the input signal, and this information can be used to create frameworks for tracking noise and conducting a priori signal to-noise ratio estimation. Example frameworks provide increased flexibility for various aspects of system design, such as gain estimation. Examples include training a DNN to detect speech in the presence of both noise and reverberation, enabling joint suppression of additive noise and reverberation. Example frameworks provide significant improvements in objective speech quality metrics relative to baseline systems.

Abstract: Apparatus for determining the location of a signal-generating source (e.g., a conferee in a telephone conference) includes at least three sensors (e.g., microphones) arranged in a plurality of sets, each having two or more sensors. A surface-finding element responds to receipt at each sensor set of signals (e.g., speech) from the source for identifying a geometric surface (e.g., the surface of a hyperboloid or cone) representing potential locations of the source as a function of sensor locations and time difference of arrival of the signals. A location-approximating element coupled to two or more of the sets identifies a line that further defines potential source locations at the intersection of the surfaces. A location signal representing those potential locations is generated in accord with parameters of that line. Further functionality generates generating the location signal as a function of closest intersections the plural ones of the aforementioned lines.

Abstract: Apparatus for determining the location of a signal-generating source (e.g., a conferee in a telephone conference) includes at least three sensors (e.g., microphones) arranged in a plurality of sets, each having two or more sensors. A surface-finding element responds to receipt at each sensor set of signals (e.g., speech) from the source for identifying a geometric surface (e.g., the surface of a hyperboloid or cone) representing potential locations of the source as a function of sensor locations and time difference of arrival of the signals. A location-approximating element coupled to two or more of the sets identifies a line that further defines potential source locations at the intersection of the surfaces. A location signal representing those potential locations is generated in accord with parameters of that line. Further functionality generates generating the location signal as a function of closest intersections the plural ones of the aforementioned lines.

Abstract: Methods and systems for beamforming are disclosed that include a signal processor that can dynamically determine the relative time delays between a plurality of frequency-dependent signals. The signal processor can adaptively generate a beam signal by aligning the plural frequency-dependent signals according to the relative time delays between the signals. The signal processor can store one frequency-dependent signal as a reference signal and can align the remaining frequency-dependent signals relative to this reference signal. One advantage of the signal processor is that it can align the plural frequency-dependent signals generated by an array of microphones that can be arranged in a linear, two dimensional or three dimensional array and located in a room environment.