DEREVERBERATION PARAMETER ESTIMATION DEVICE AND METHOD, DEREVERBERATION/ECHO-CANCELLATION PARAMETERESTIMATIONDEVICE,DEREVERBERATIONDEVICE,DEREVERBERATION/ECHO-CANCELLATION DEVICE, AND DEREVERBERATION DEVICE ONLINE CONFERENCING SYSTEM

• US 20150016622A1
• Filed: 02/15/2013
• Published: 01/15/2015
• Est. Priority Date: 02/17/2012
• Status: Active Grant

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Togami, Masahito, Kawaguchi, Yohei
• Granted Patent: US 9,288,576 B2
• Time in Patent Office: Days
• US Class Current: 381/66
• CPC Class Codes: G10H 2210/281 Reverberation or echo
• CPC Class Codes: G10L 2021/02082 the noise being echo, reverberation of the speech
• CPC Class Codes: G10L 21/0208 Noise filtering
• CPC Class Codes: H04B 3/23 using a replica of transmitted signal in the time domain, e.g. echo cancellers
• CPC Class Codes: H04M 9/082 using echo cancellers echo cancellers per se H04B3/23
• CPC Class Codes: H04R 2430/03 Synergistic effects of band splitting and sub-band processing
• CPC Class Codes: H04R 3/002 Damping circuit arrangements for transducers, e.g. motional feedback circuits
• CPC Class Codes: H04R 3/005 for combining the signals of two or more microphones specially adapted for hearing aids H04R25/407
• CPC Class Codes: H04R 3/02 for preventing acoustic reaction , i.e. acoustic oscillatory feedback specially adapted for hearing aids H04R25/453

Abstract

In a conventional dereverberation system, when there is a fluctuating reverberation component, it has been difficult to determine, with high accuracy, a linear dereverberation filter for removing a non-fluctuating reverberation component. An algorithm integrating a dereverberation system using a linear filter and a dereverberation system using a non-linear filter includes the function of measuring the amount of fluctuation in transfer function in a latter-stage non-linear filter over time, and controls the strength of the non-linear filter over time based on the function. In this configuration, a strong non-linear process is implemented only when the fluctuation in transfer function is large, whereby distortion in speech components can be minimized. Further, by feeding the amount of fluctuation in transfer function over time by the function back to the estimation of a previous-stage linear filter, and thus reducing the weight for a time band in which the transfer function of a linear filter estimated value fluctuates, the influence causing a decrease in estimation accuracy of the transfer function can be reduced.

Claims

1. A dereverberation parameter estimation device comprising:

a memory storing a parameter of a linear filter for removing a non-fluctuating reverberation component contained in a speech input signal, and a parameter of a non-linear filter for removing a fluctuating reverberation component contained in the speech input signal; and

a processor that estimates a dereverberation parameter for acquiring a direct sound by removing the reverberation components contained in the speech input signal, and that stores the dereverberation parameter in the memory as the parameter of the linear filter and the parameter of the non-linear filter,

wherein the processor is configured to;

read the parameter of the linear filter from the memory, and generates an estimated value of a linear dereverberation signal by removing the non-fluctuating reverberation component contained in the speech input signal using the linear filter;

read from the memory the parameter of the non-linear filter, and generates estimated values of the fluctuating reverberation component and the direct sound component contained in the estimated value of the linear dereverberation signal using the non-linear filter;

based on the estimated values of the fluctuating reverberation component and the direct sound, update the parameters of the variation reverberation component and the direct sound component constituting the parameter of the non-linear filter; and

successively update the parameter of the linear filter based on the parameters of the updated fluctuating reverberation component and direct sound component.

2. The dereverberation parameter estimation device according to claim 1, wherein the processor, using the updated parameters of the fluctuating reverberation component and the direct sound component, determines a weight coefficient for the linear filter by reflecting a time band having a large sum of the variation reverberation component and the direct sound component in the speech input signal, and updates the parameter of the linear filter.

3. The dereverberation parameter estimation device according to claim 1,

wherein the speech input signal includes speech signals from a plurality of microphones,

wherein the processor is configured to:

generate the estimated value of the direct sound component and the estimated value of the fluctuating reverberation component contained in the speech signal from each of the plurality of microphones,

extract a secondary statistical amount of the estimated values of the direct sound component and the fluctuating reverberation component from each microphone; and

use, as the parameters of the fluctuating reverberation component and the direct sound component, information from the secondary statistical amount indicating a dispersion of power of each of the direct sound component and the fluctuating reverberation component of the speech signals from the plurality of microphones.

4. The dereverberation parameter estimation device according to claim 1, wherein the processor is configured to execute a convergence determination on the updated parameters of the fluctuating reverberation component and the direct sound component,

wherein the processor is configured to output the parameters of the linear filter and the non-linear filter as the dereverberation parameter if it is determined that the parameters have converged, or continues the successive updating process of the parameter of the linear filter if it is determined that the parameters have not converged.

8. A dereverberation device for removing a reverberation component contained in a speech input signal, comprising:

the dereverberation parameter estimation device according to claim 1; and

a processor that acquires from the dereverberation parameter estimation device the dereverberation parameter, and that outputs the direct sound component by removing from the speech input signal the non-fluctuating reverberation component and the fluctuating reverberation component,

wherein the processor is configured to;

estimate a linear dereverberation signal by removing the non-fluctuating reverberation component from the speech input signal using the linear filter,

separate the fluctuating reverberation component and the direct sound component from the linear dereverberation signal using the non-linear filter, and

learn a change in power of the direct sound component over time and outputs the direct sound component in which the change in the power has converged.

9. A dereverberation device for removing a reverberation component contained in a speech input signal, comprising:

the dereverberation parameter estimation device according to claim 1;

a dereverberation parameter database storing the dereverberation parameter obtained by the dereverberation parameter estimation device in the past; and

a processor that is configured to;

acquire from the dereverberation parameter database a plurality of dereverberation parameters, remove the non-fluctuating reverberation component and the fluctuating reverberation component from the speech input signal using each of the plurality of dereverberation parameters, generate a plurality of direct sound components, and output an optimum direct sound component among the plurality of direct sound components.

10. The dereverberation device according to claim 9, wherein the processor, using each of the plurality of dereverberation parameters, is configured to:

estimate a linear dereverberation signal by removing, using the linear filter, the non-fluctuating reverberation component from the speech input signal;

separate the fluctuating reverberation component and the direct sound component from the linear dereverberation signal using the non-linear filter; and

learn a change in power of the direct sound component over time, and outputs the direct sound component in which the change in power has converged in accordance with the plurality of dereverberation parameters.

12. An on-line conferencing system comprising:

a plurality of hub-based conferencing systems disposed at respective hubs; and

a computation server that manages the plurality of hub-based conferencing systems,

wherein the computation server includes the dereverberation parameter estimation device according to claim 1, processes the speech input signal received from each hub-based conferencing system using the dereverberation parameter estimation device, computes a dereverberation parameter corresponding to each hub-based conferencing system, and transmits the computed reverberation parameter to a corresponding hub-based conferencing system,

wherein each of the plurality of hub-based conferencing systems transmits the speech input signal to the computation server, receives from the computation server the dereverberation parameter corresponding to the transmitted speech input signal, separates the non-fluctuating reverberation component and the fluctuating reverberation component from the speech input signal using the received reverberation parameter, and outputs the direct sound component in real-time.

5. A reverberation/echo removal parameter estimation device comprising:

a memory storing a parameter of a linear reverberation filter for removing a non-fluctuating reverberation component contained in a speech input signal, a parameter of a linear echo erasing filter for removing a non-fluctuating echo component contained in the speech input signal, and a parameter of a non-linear filter for removing a fluctuating reverberation component and a fluctuating echo component contained in the speech input signal; and

a processor that estimates a reverberation/echo removal parameter for acquiring a direct sound by removing the reverberation component and the echo component contained in the speech input signal, and that stores the reverberation/echo removal parameter in the memory as the parameter of the linear reverberation filter, the parameter of the linear echo erasing filter, and the parameter of the non-linear filter,

wherein the processor is configured to;

read from the memory the parameter of the linear reverberation filter and the parameter of the linear echo erasing filter, and generates an estimated value of a linear reverberation/echo removal signal by removing the non-fluctuating reverberation component and the non-fluctuating echo component contained in the speech input signal using the linear reverberation filter and the linear echo erasing filter,

read from the memory the parameter of the non-linear filter, and generates estimated values of the fluctuating reverberation component, the fluctuating echo component, and the direct sound component contained in the estimated value of the linear reverberation/echo removal signal, using the non-linear filter,

update the parameters of the variation reverberation component, the fluctuating echo component, and the direct sound component constituting the parameter of the non-linear filter based on the estimated values of the fluctuating reverberation component, the fluctuating echo component, and the direct sound, and

successively update the parameters of the linear reverberation filter and the linear echo erasing filter based on the updated parameters of the fluctuating reverberation component, the fluctuating echo component, and the direct sound component.

6. The reverberation/echo removal parameter estimation device according to claim 5, wherein the processor is configured to update the parameter of the linear echo erasing filter applied to the speech signal after the fluctuating reverberation component is removed, and update the parameter of the linear dereverberation filter applied to the speech signal after the fluctuating echo component is removed.

11. A reverberation/echo removal device for removing a reverberation component contained in a speech input signal, comprising:

the reverberation/echo removal parameter estimation device according to claim 5; and

a processor that acquires from the reverberation/echo removal parameter estimation device the reverberation/echo removal parameter, that removes the non-fluctuating and fluctuating reverberation components and non-fluctuating and fluctuating echo components from the speech input signal, and that outputs the direct sound component,

wherein the processor is configured to;

estimate a linear reverberation/echo removal signal by removing the non-fluctuating reverberation component and the non-fluctuating echo component from the speech input signal using the linear reverberation filter and the linear echo erasing filter,

separate the fluctuating reverberation and echo components and the direct sound component from the linear reverberation/echo removal signal using the non-linear filter, and

learn a change in power of the direct sound component over time, and outputs the direct sound component in which the change in power has converged.

7. A dereverberation parameter estimation method including, using a processor, estimating a dereverberation parameter for acquiring a direct sound by removing a reverberation component contained in a speech input signal, and outputting the dereverberation parameter as a parameter of a linear filter for removing a non-fluctuating reverberation component contained in the speech input signal, and as a parameter of a non-linear filter for removing a fluctuating reverberation component contained in the speech input signal,

the method comprising the steps of:

the processor reading from the memory the parameter of the linear filter, and generating an estimated value of a linear dereverberation signal by removing the non-fluctuating reverberation component contained in the speech input signal using the linear filter;

the processor reading from the memory the parameter of the non-linear filter, and generating estimated values of the fluctuating reverberation component and the direct sound component contained in the estimated value of the linear dereverberation signal using the non-linear filter;

the processor, based on the estimated values of the fluctuating reverberation component and the direct sound, updating the parameters of the variation reverberation component and the direct sound component constituting the parameter of the non-linear filter; and

the processor, based on the updated parameters of the fluctuating reverberation component and the direct sound component, successively updating the parameter of the linear filter.