Active noise control system

Abstract

To provide an "active noise control system" that can cancel noise successfully in a high frequency range in a blind type configuration where noise is not detected upstream.SOLUTION: An update control section 16 causes an ICA processing section 15 to perform independent component analysis processing when a sound level output by a microphone 11 becomes great. In the independent component analysis, the ICA processing section 15 separates a noise component from observed sound data output by the microphone 11 by independent component analysis, and stores the data representing the separated noise component as statistical noise data 14. An adaptive filter 13, which performs adaptive operation using the output of the microphone 11 as an error, and outputs, as a noise cancellation sound NC from a speaker 12, a sound represented by the noise cancellation sound data, which is obtained by filtering the stored statistical noise data 14. The adaptive filter 13 also adapts the filter coefficients of the filter processing using the observed sound data output by the microphone 11 as an error.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to an active noise control (ANC) technique that reduces noise by using a noise canceling sound that cancels the noise. [Background technology]

[0002] Known active noise control techniques include feedforward active noise control and feedback active noise control (for example, Patent Document 1). In active noise control, a sensor detects events that are correlated with noise, such as vibrations and rotations of the mechanism that is the noise source at a position upstream of the target position for noise cancellation, and sounds near the noise source, to generate a reference signal. An adaptive filter that uses the sound picked up by a microphone placed near the target position for noise cancellation as an error signal generates a noise cancellation sound that cancels out the noise from the reference signal and outputs it from a speaker.

[0003] On the other hand, in feedback-type active noise control, a noise-canceling sound is also generated by an adaptive filter and output from a speaker. However, in feedback-type active noise control, a sound picked up by a microphone placed near the position to be canceled is used as an error signal, and a signal obtained by adding an estimated noise-canceling sound propagating from the speaker to the microphone to the error signal is used as a noise signal representing noise, and a noise-canceling sound that cancels the noise from the noise signal is generated by an adaptive filter adapted to the error signal.

[0004] In addition, known technologies related to the present application include a technology for separating a mixture of speech and noise into speech and noise using independent component analysis (ICA) (e.g., Patent Document 2), and a technology for separating individual audio sources from an audio signal containing a mixture of multiple audio sources using a neural network (e.g., Patent Document 3).

[0005] Also, as a technology related to the present application, a technology is known in which the first and second principal components of a sound are obtained by principal component analysis using the intensity of each frequency band of the sound as data (variable) (for example, Patent Document 4). [Prior art documents] [Patent documents]

[0006] [Patent Document 1] JP 2017-090736 A [Patent Document 2] Special Publication No. 2006-510069 [Patent Document 3] Special Publication No. 2009-511954 [Patent Document 4] Japanese Patent Application Publication No. 11-117875 Summary of the Invention [Problem to be solved by the invention]

[0007] The above-mentioned feedforward type active noise control requires, in addition to a microphone that picks up error for adapting an adaptive filter placed near the position to be canceled, a special sensor that detects events that are correlated with noise at a position upstream of the position to be canceled in the noise flow in order to generate a reference signal; if, for various reasons, it is not possible to provide such a sensor, active noise control cannot be performed.

[0008] On the other hand, feedback-type active noise control does not require any sensors to detect events correlated with noise other than a microphone placed near the location where noise is to be canceled, but it has the problem that it is difficult to cancel noise in the high-frequency range due to propagation delays in the feedback loop. Also, in feedforward active noise control, if a sensor that detects an event correlated with noise is close to the position to be canceled, there is insufficient calculation processing time to generate the required high-frequency components of the noise-canceling sound, making it difficult to effectively cancel noise in the high-frequency range.

[0009] Therefore, an object of the present invention is to realize active noise control that can effectively cancel noise even in the high frequency range in a blind type configuration that does not have a sensor that detects events correlated with noise at a position upstream of the noise cancellation target position. [Means for solving the problem]

[0010] In order to achieve the above object, the present invention provides an active noise control system that performs active noise control, comprising a microphone that outputs a picked up sound as an observation signal, a speaker that outputs a noise-canceling sound, an adaptive filter, a storage means, a noise separation means that performs a noise separation process at a predetermined opportunity to separate noise components from the observation signal and store a pseudo-noise signal representing the separated noise components in the storage means, and an adaptive filter that receives as input the pseudo-noise signal stored in the storage means, wherein the adaptive filter performs an adaptive operation using the observation signal as an error and generates a noise-canceling sound to be output from the speaker from the pseudo-noise signal.

[0011] In order to achieve the above object, the present invention provides an active noise control system that performs active noise control, comprising a microphone, frequency analysis means for converting an output of the microphone into a frequency domain signal and outputting it as an observation signal, a speaker for outputting a noise-canceling sound, an adaptive filter, storage means, time domain conversion means, noise separation means for performing a noise separation process at a predetermined opportunity, which separates noise components from the observation signal in the frequency domain and stores a pseudo-noise signal representing the separated noise components in the frequency domain in the storage means, and an adaptive filter that receives as input the pseudo-noise signal stored in the storage means, wherein the adaptive filter performs an adaptive operation in the frequency domain using the observation signal as an error, and the time domain conversion means converts the output of the adaptive filter into a time domain signal to generate the noise-canceling sound to be output from the speaker.

[0012] Here, this active noise control system may be configured so that the frequency analysis means converts the output of the microphone into a frequency domain signal by fast Fourier transform and outputs it as an observation signal, and the time domain conversion means converts the output of the adaptive filter into a time domain signal by inverse fast Fourier transform.

[0013] In order to achieve the above object, the present invention provides an active noise control system that performs active noise control, comprising a microphone, analysis means for converting an output of the microphone into weighting coefficients of a predetermined transformation function that generates a time domain signal by weighting an orthogonal basis and outputting the weighted output as an observation signal, a speaker that outputs a noise-canceled sound, an adaptive filter, storage means, time domain transformation means, noise separation means that performs a noise separation process at a predetermined opportunity to separate a noise component from the observation signal and store the noise component in the storage means as a pseudo-noise signal, and an adaptive filter that receives as input the pseudo-noise signal stored in the storage means, wherein the adaptive filter performs an adaptive operation using the observation signal as an error, and the time domain transformation means transforms the output of the adaptive filter into a time domain signal using the transformation function to generate a noise-canceled sound to be output from the speaker.

[0014] Here, in this active noise control system, the analysis means may convert the output of the microphone into the weighting coefficient by fast Fourier transform and output it as an observation signal, and the time domain transformation means may convert the output of the adaptive filter into a time domain signal by inverse fast Fourier transform.

[0015] Alternatively, the analysis means may convert the microphone output into the weighting coefficients by wavelet transform and output the weighting coefficients as observed signals, and the time domain conversion means may convert the adaptive filter output into a time domain signal by inverse wavelet transform. The above active noise control system may be configured so that the noise separating means performs the noise separating process when the output of the microphone exceeds a predetermined level. Moreover, the above active noise control system may be configured such that the noise separating means separates noise components from the observed signal by independent component analysis (ICA). Alternatively, the above-described active noise control system may be configured such that the noise separating means infers and separates noise components from the observation signals using a neural network that has been trained in advance to infer noise components from the observation signals. Alternatively, the above active noise control system may be configured such that the noise separating means calculates principal components of a predetermined order of the observed signal by principal component analysis, and separates the calculated principal components as the noise components. According to the above-described active noise control system, the noise components in the output of the microphone that picks up the error for adapting the adaptive filter are separated and stored as a pseudo noise signal, and the stored pseudo noise signal is used as the input of the adaptive filter to generate a noise-canceled sound by active noise control. Also, unlike feedback-type active noise control, there is no problem of delay in feedback loop propagation that causes noise to be unable to be cancelled satisfactorily in the high-frequency range. Furthermore, since pre-stored pseudo noise data is used as the input of the adaptive filter, there is no problem in feedforward-side active noise control where a sensor that detects an event correlated with noise at a position upstream of the noise cancellation target position is close to the noise cancellation target position, and therefore there is no problem of noise being unable to be cancelled satisfactorily in the high-frequency range due to insufficient calculation processing time. Therefore, according to the present invention, in a blind type configuration in which a sensor for detecting an event correlated with noise is not provided at a position upstream of the noise cancellation target position, active noise control can be performed that can effectively cancel noise even in the high frequency range.

[0016] In order to achieve the above object, the present invention provides an active noise control system that performs active noise control, comprising a microphone, analysis means for converting an output of the microphone into a weighting coefficient of a predetermined transformation function that generates a time domain signal by weighting an orthogonal basis and outputting the weighted output as an observation signal, a speaker for outputting a noise-canceling sound, an adaptive filter, storage means for storing in advance, as a pseudo-noise signal, the weighting coefficient for generating a pseudo-noise by the transformation function, time domain transformation means, and an adaptive filter that receives as input the pseudo-noise signal stored in the storage means, the adaptive filter performing an adaptive operation using the observation signal as an error, and the time domain transformation means for transforming the output of the adaptive filter into a time domain signal by the transformation function to generate the noise-canceling sound to be output from the speaker.

[0017] Here, in this active noise control system, the analysis means may convert the output of the microphone into the weighting coefficient by fast Fourier transform and output it as an observation signal, and the time domain transformation means may convert the output of the adaptive filter into a time domain signal by inverse fast Fourier transform.

[0018] Alternatively, the analysis means may convert the microphone output into the weighting coefficients by wavelet transform and output the weighting coefficients as observed signals, and the time domain conversion means may convert the adaptive filter output into a time domain signal by inverse wavelet transform. Such an active noise control system can also perform active noise control that can effectively cancel noise even in the high frequency range in a blind type configuration that does not have a sensor that detects events correlated with noise at a position upstream of the noise cancellation target position. Effect of the Invention

[0019] As described above, according to the present invention, it is possible to realize active noise control that can effectively cancel noise even in the high frequency range in a blind type configuration that does not have a sensor for detecting an event correlated with noise at a position upstream of the noise cancellation target position. [Brief description of the drawings]

[0020] [Figure 1] 1 is a block diagram showing a configuration of a noise control device according to a first embodiment of the present invention. [Diagram 2] FIG. 5 is a block diagram showing the configuration of a noise control device according to a second embodiment of the present invention. [Diagram 3] FIG. 11 is a block diagram showing the configuration of a noise control device according to a third embodiment of the present invention. [Figure 4] FIG. 11 is a block diagram showing the configuration of a noise control device according to a fourth embodiment of the present invention. [Diagram 5] FIG. 11 is a block diagram showing the configuration of a noise control device according to a fifth embodiment of the present invention. [Figure 6] FIG. 13 is a block diagram showing the configuration of a noise control device according to a sixth embodiment of the present invention. [Figure 7] FIG. 13 is a block diagram showing the configuration of a noise control device according to a seventh embodiment of the present invention. [Figure 8] FIG. 13 is a block diagram showing the configuration of a noise control device according to an eighth embodiment of the present invention. [Figure 9] FIG. 4 is a block diagram illustrating another configuration example of a noise control device according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0021] Hereinafter, an embodiment of the present invention will be described. First, the first embodiment will be described. FIG. 1 shows the configuration of a noise control device according to the first embodiment. As shown in the figure, the noise control device includes a microphone 11 placed near the position to be canceled, a speaker 12 that outputs a noise cancellation sound NC that cancels noise N at the position to be canceled, an adaptive filter 13, statistical noise data 14, an ICA processing unit 15 that performs independent component analysis (ICA), and an update control unit 16.

[0022] The microphone 11 outputs a mixed sound of the picked up noise N and the noise canceling sound NC as observed sound data. Here, when the noise canceling sound NC is being output from the speaker 12, this mixed sound data represents the remaining components of the noise N that could not be completely canceled by the noise canceling sound NC. The update control unit 16 stops the operation of the adaptive filter 13 when setting up the noise control device or when the sound level represented by the observed sound data output by the microphone 11 becomes larger than a predetermined level, and causes the ICA processing unit 15 to perform independent component analysis processing. Here, during the period when the operation of the adaptive filter 13 is stopped, the adaptive operation of the adaptive filter 13 and the output of the noise cancellation sound NC are stopped. Therefore, during the period when the operation of the adaptive filter 13 is stopped, the observed sound data output by the microphone 11 represents the noise N. Now, in the independent component analysis process, the ICA processing unit 15 separates noise components from the observed sound data output by the microphone 11 based on statistical independence through independent component analysis, and stores data representing the separated noise components as statistical noise data 14. Then, once the statistical noise data 14 has been stored, the update control unit 16 resumes the operation of the adaptive filter 13 . The stored statistical noise data 14 is output to the adaptive filter 13 as pseudo noise data. The adaptive filter 13 outputs a sound represented by noise-canceled sound data obtained by filtering the input pseudo noise data as a noise-canceled sound NC from the speaker 12. The adaptive filter 13 also treats the observed sound data output by the microphone 11 as an error and adapts the filter coefficients of the filtering process applied to the pseudo noise data by an adaptive algorithm such as LMS or NLMS so as to minimize the power of the error.

[0023] A second embodiment of the present invention will now be described. In the second embodiment, signal processing equivalent to the operation of the noise control device of the first embodiment is performed in the frequency domain. FIG. 2 shows the configuration of a noise control device according to the second embodiment. As shown in the figure, the noise control device includes a microphone 21 placed near the position to be canceled, a speaker 22 that outputs a noise cancellation sound NC that cancels noise N at the position to be canceled, an adaptive filter 23, frequency domain statistical noise data 24, an ICA processing unit 25 that performs independent component analysis (ICA), an update control unit 26, a frequency analysis unit 27, and a time domain conversion unit 28.

[0024] The frequency analysis unit 27 performs frequency analysis on the observation sound data output by the microphone 21, and converts the observation sound data into frequency domain observation sound data that represents the sound intensity for each frequency band. For example, FFT (Fast Fourier Transform) can be used as the frequency analysis performed by the frequency analysis unit 27. When FFT is used, the frequency domain observed sound data becomes a frequency spectrum. The update control unit 26 stops the operation of the adaptive filter 23 and causes the ICA processing unit 25 to perform independent component analysis processing when setting up the noise control device or when it is detected from the frequency domain observation sound data that the sound level represented by the observation sound data output by the microphone 21 has become greater than a predetermined level.

[0025] In the independent component analysis process, the ICA processing unit 25 performs independent component analysis in the frequency domain, separates data representing noise components in the frequency domain from the frequency domain observed sound data based on statistical independence, and stores the separated data as frequency domain statistical noise data 24. Here, when FFT is used as the frequency analysis performed by the frequency analysis unit 27, the frequency domain statistical noise data 24 also becomes a frequency spectrum. Once the statistical noise data 14 has been stored, the update control unit 26 resumes the operation of the adaptive filter 23 . The stored frequency domain statistical noise data 24 is output to the adaptive filter 23 as frequency domain pseudo noise data. The adaptive filter 23 performs filtering on the input frequency domain artificial noise data in the frequency domain and outputs the result as frequency domain noise-canceled sound data to the time domain transform section 28. The adaptive filter 23 also treats the frequency domain observed sound data output by the frequency analysis section 27 as an error and adapts the filter coefficients of the filtering performed on the frequency domain artificial noise data by an adaptive algorithm such as LMS or NLMS so as to minimize the power of the error.

[0026] The time domain conversion unit 28 converts the frequency domain noise-canceled sound data output by the adaptive filter 23 into a sound in the time domain, and outputs the sound from the speaker 22 as the noise-canceled sound NC. Here, when FFT is used as the frequency analysis performed by frequency analysis unit 27, the transformation performed by time domain transformation unit 28 is IFFT (inverse fast Fourier transform). A third embodiment of the present invention will now be described. In the third embodiment, the noise control device of the first embodiment separates noise from observed sound data using a neural network instead of independent component analysis (ICA). FIG. 3 shows the configuration of a noise control device according to the third embodiment. As shown in the figure, the noise control device includes a microphone 31 placed near the position to be canceled, a speaker 32 that outputs a noise cancellation sound NC that cancels noise N at the position to be canceled, an adaptive filter 33, inferred noise data 34, a trained neural network 35, and an update control unit 36.

[0027] The microphone 31 outputs a mixed sound of the picked up noise N and the noise canceling sound NC as observed sound data. The update control unit 36 ​​stops the operation of the adaptive filter 33 when the noise control device is set up or when the sound level represented by the observed sound data output by the microphone 31 becomes greater than a predetermined level, and causes the trained neural network 35 to perform noise inference processing.

[0028] The trained neural network 35 is a neural network that has been trained in advance to perform inference to separate noise components from observed sound data using appropriate teacher data. In the noise inference process, the network infers noise from the observed sound data output by the microphone 31, and stores data representing the inferred noise as inferred noise data 34.

[0029] Once the inferred noise data 34 has been stored, the update control unit 36 ​​resumes the operation of the adaptive filter 33 . The stored inferred noise data 34 is output to the adaptive filter 33 as pseudo noise data. The adaptive filter 33 outputs a sound represented by noise-canceled sound data obtained by filtering the input pseudo noise data from the speaker as noise-canceled sound NC. The adaptive filter 33 also treats the observed sound data output by the microphone 31 as an error and adapts the filter coefficients of the filtering process applied to the pseudo noise data using an adaptive algorithm such as LMS or NLMS so as to minimize the power of the error.

[0030] The fourth embodiment of the present invention will now be described. In the fourth embodiment, signal processing equivalent to the operation of the noise control device of the fourth embodiment is performed in the frequency domain. FIG. 4 shows the configuration of a noise control device according to the fourth embodiment. As shown in the figure, the noise control device includes a microphone 41 arranged near the position to be canceled, a speaker 42 that outputs a noise cancellation sound NC that cancels noise N at the noise cancellation target position, an adaptive filter 43, frequency domain inference noise data 44, a trained neural network 45, an update control unit 46, a frequency analysis unit 47, and a time domain conversion unit 48.

[0031] The frequency analysis unit 47 performs frequency analysis on the observation sound data output by the microphone 41, and converts the observation sound data into frequency domain observation sound data that represents the sound intensity for each frequency band. For example, FFT (Fast Fourier Transform) can be used as the frequency analysis performed by the frequency analysis unit 47. When FFT is used, the frequency domain observed sound data becomes a frequency spectrum. The update control unit 46 stops the operation of the adaptive filter 43 and causes the trained neural network 45 to perform noise inference processing when setting up the noise control device or when it is detected from the frequency domain observation sound data that the sound level represented by the observation sound data output by the microphone 41 has exceeded a predetermined level.

[0032] The trained neural network 45 is a neural network that has been trained in advance using appropriate teacher data to perform inference to separate noise components from frequency domain observed sound data in the frequency domain. In the noise inference process, the noise components are inferred from the frequency domain observed sound data in the frequency domain, and data representing the noise components in the inferred frequency domain is stored as frequency domain inferred noise data 44.

[0033] Here, when FFT is used as the frequency analysis performed by the frequency analysis unit 47, the frequency domain inference noise data 44 also becomes a frequency spectrum. Once the frequency domain inference noise data 44 has been stored, the update control unit 46 resumes the operation of the adaptive filter 43 . The stored frequency domain inferred noise data 44 is output to the adaptive filter 43 as frequency domain pseudo noise data. The adaptive filter 43 performs filtering on the input frequency domain artificial noise data in the frequency domain and outputs the result as frequency domain noise-canceled sound data to the time domain conversion unit 48. The adaptive filter 43 also treats the frequency domain observed sound data output by the frequency analysis unit 47 as error and adapts the filter coefficients of the filtering performed on the frequency domain artificial noise data by an adaptive algorithm such as LMS or NLMS so as to minimize the power of the error.

[0034] The time domain converter 48 converts the frequency domain noise-canceled sound data output by the adaptive filter 43 into a time domain sound and outputs it from the speaker 42 as a noise-canceled sound NC. Here, when FFT is used as the frequency analysis performed by the frequency analysis unit 47, the transformation performed by the time domain transformation unit 48 is IFFT (inverse fast Fourier transform). The fifth embodiment of the present invention will now be described. In the fifth embodiment, the noise control device of the first embodiment separates noise from observed sound data using principal component analysis instead of independent component analysis (ICA). FIG. 5 shows the configuration of a noise control device according to the fifth embodiment. As shown in the figure, the noise control device includes a microphone 51 placed near the position to be canceled, a speaker 52 that outputs a noise cancellation sound NC that cancels noise N at the position to be canceled, an adaptive filter 53, principal component data 54, a principal component analysis unit 55, and an update control unit 56.

[0035] The microphone 51 outputs a mixed sound of the picked up noise N and the noise canceling sound NC as observed sound data. The update control unit 56 stops the operation of the adaptive filter 53 and causes the principal component analysis unit 55 to perform principal component analysis when setting up the noise control device or when the sound level represented by the observed sound data output by the microphone 51 becomes greater than a predetermined level. The principal component analysis unit 55 analyzes principal components of a predetermined order of the observed sound data (for example, only the first principal component, or only the second principal component, or the first principal component and the second principal component) by principal component analysis, and stores data representing the sound for which the cumulative contribution rate of each of the analyzed principal components is maximum as principal component data 54.

[0036] Once the principal component data 54 has been stored, the update control unit 56 resumes the operation of the adaptive filter 53 . The stored principal component data 54 is output to the adaptive filter 53 as pseudo noise data. The adaptive filter 53 outputs a sound represented by noise-canceled sound data obtained by filtering the input pseudo-noise data as a noise-canceled sound NC from the speaker 52. The adaptive filter 53 also treats the observed sound data output by the microphone 51 as an error and adapts the filter coefficients of the filtering process applied to the pseudo-noise data by an adaptive algorithm such as LMS or NLMS so as to minimize the power of the error.

[0037] Next, a sixth embodiment of the present invention will be described. In the sixth embodiment, signal processing equivalent to the operation of the noise control device in the fifth embodiment is performed in the frequency domain. FIG. 6 shows the configuration of a noise control device according to the sixth embodiment. As shown in the figure, the noise control device includes a microphone 61 placed near the position to be canceled, a speaker 62 that outputs a noise cancellation sound NC that cancels noise N at the position to be canceled, an adaptive filter 63, frequency domain principal component data 64, a principal component analysis unit 65, an update control unit 66, a frequency analysis unit 67, and a time domain conversion unit 68.

[0038] The frequency analysis unit 67 performs frequency analysis on the observation sound data output by the microphone 61, and converts the observation sound into frequency domain observation sound data that represents the sound intensity for each frequency band. For example, FFT (Fast Fourier Transform) can be used as the frequency analysis performed by the frequency analysis unit 67. When FFT is used, the frequency domain observed sound data becomes a frequency spectrum. The update control unit 66 stops the operation of the adaptive filter 63 and causes the principal component analysis unit 65 to perform principal component analysis when setting up the noise control device or when it is detected from the frequency domain observation sound data that the sound level represented by the observation sound data output by the microphone 61 has become greater than a predetermined level.

[0039] The principal component analysis unit 65 analyzes principal components of a predetermined order (for example, only the first principal component, or only the second principal component, or the first principal component and the second principal component) of the frequency domain observed sound data by principal component analysis in the frequency domain, and stores data representing the sound in which the cumulative contribution rate of each analyzed principal component is maximum in the frequency domain as frequency domain principal component data 64. Here, when FFT is used as the frequency analysis performed by the frequency analysis unit 67, the frequency domain principal component data 64 also becomes a frequency spectrum.

[0040] Once the frequency domain principal component data 64 has been stored, the update control unit 66 resumes the operation of the adaptive filter 63 . The stored frequency domain principal component data 64 is output to the adaptive filter 63 as frequency domain pseudo noise data. The adaptive filter 63 performs filtering on the input frequency domain artificial noise data in the frequency domain and outputs the result as frequency domain noise-canceled sound data to the time domain conversion unit 68. The adaptive filter 63 also treats the frequency domain observed sound data output by the frequency analysis unit 67 as error and adapts the filter coefficients of the filtering performed on the frequency domain artificial noise data using an adaptive algorithm such as LMS or NLMS so as to minimize the power of the error.

[0041] The time domain conversion unit 68 converts the frequency domain noise-canceled sound data output by the adaptive filter 63 into a sound in the time domain, and outputs the sound from the speaker 62 as the noise-canceled sound NC. Here, when FFT is used as the frequency analysis performed by the frequency analysis unit 67, the transformation performed by the time domain transformation unit 68 is IFFT (inverse fast Fourier transform). The seventh embodiment of the present invention will now be described. FIG. 7 shows the configuration of a noise control device according to the seventh embodiment. As shown in the figure, the noise control device includes a microphone 701 placed near the noise cancellation target position, a speaker 702 that outputs a noise cancellation sound NC that cancels noise N at the noise cancellation target position, an adaptive filter 703, a weighting coefficient 704, an ICA processing unit 705 that performs independent component analysis (ICA), an update control unit 706, a frequency analysis unit 707, a time domain transformation unit 708, a basis spectrum calculation unit 709, and orthogonal basis data 710.

[0042] The orthogonal basis data 710 is data of an orthogonal basis used in a transformation function that generates a time domain signal by weighting a predetermined orthogonal basis with a weight. As such a transformation function, an inverse fast Fourier transform or the like can be used. When an inverse fast Fourier transform is used, the orthogonal basis data 710 becomes data of a Fourier basis.

[0043] The frequency analysis unit 707 performs frequency analysis on the observation sound data output by the microphone 701, and converts the observation sound data into frequency domain observation sound data representing the weighting coefficients of the orthogonal basis of the above-mentioned conversion function. Here, when the orthogonal basis data 710 is Fourier basis data, the frequency analysis performed by the frequency analysis unit 707 uses FFT (Fast Fourier Transform), and the frequency domain observed sound data becomes a frequency spectrum. The update control unit 706 stops the operation of the adaptive filter 703 and causes the ICA processing unit 705 to perform independent component analysis processing when setting up the noise control device or when it is detected from the frequency domain observation sound data that the level of the sound represented by the observation sound data output by the microphone 701 has increased to a predetermined level or more.

[0044] In the independent component analysis process, the ICA processing unit 705 performs independent component analysis in the frequency domain, separates data representing frequency domain noise components from the frequency domain observed sound data based on statistical independence, and sends the separated data to the basis spectrum calculation unit 709 as frequency domain statistical noise data.

[0045] The basis spectrum calculation unit 709 calculates a complex spectrum for each element of the orthogonal basis stored in the orthogonal basis data 710, of the noise represented in the frequency domain by the frequency domain statistical noise data sent from the ICA processing unit 705, and sets the weighting coefficient of the orthogonal basis obtained from the calculated complex spectrum as the weighting coefficient 704.

[0046] The complex spectrum calculated by the basis spectrum calculation unit 709 represents a gain and a phase, and the weighting coefficients stored in the weighting coefficients 704 represent the gain and phase to be applied to each element of the orthogonal basis in complex form. Furthermore, when FFT is used as the frequency analysis performed by the frequency analysis unit 707, the weighting coefficients stored in the weighting coefficient 704 become a frequency spectrum. Once the weighting coefficients 704 have been stored, the update control unit 706 resumes the operation of the adaptive filter 703 . The stored weighting coefficients 704 are output to the adaptive filter 703 as frequency domain pseudo noise data representing the weighting coefficients of the orthogonal basis stored in the orthogonal basis data 710 . The adaptive filter 703 performs filtering on the input frequency domain artificial noise data in the frequency domain and outputs the result as frequency domain noise-canceled sound data to the time domain transform section 708. The adaptive filter 703 also treats the frequency domain observed sound data output by the frequency analysis section 707 as error and adapts the filter coefficients of the filtering performed on the frequency domain artificial noise data using an adaptive algorithm such as LMS or NLMS so as to minimize the power of the error.

[0047] The time domain transform section 708 transforms the frequency domain noise-canceled sound data output by the adaptive filter 703 into a sound in the time domain, and outputs the sound from the speaker 702 as a noise-canceled sound NC. Here, when FFT is used as the frequency analysis performed by frequency analysis unit 707, the transformation performed by time domain transformation unit 708 is IFFT (inverse fast Fourier transform). In the seventh embodiment described above, instead of the ICA processing unit 15, a processing unit may be provided that separates data representing noise in the frequency domain from the frequency domain observed sound data by using a trained neural network or principal component analysis, and converts the separated data into frequency domain statistical noise data.

[0048] The eighth embodiment of the present invention will now be described. FIG. 8 shows the configuration of a noise control device according to the eighth embodiment. As shown in the figure, the noise control device includes a microphone 801 placed near the noise cancellation target position, a speaker 802 that outputs a noise cancellation sound NC that cancels noise N at the noise cancellation target position, an adaptive filter 803, a weighting coefficient 804, an ICA processing unit 805 that performs independent component analysis (ICA), an update control unit 806, a time-frequency analysis unit 807, a time domain conversion unit 808, a basis spectrum calculation unit 809, and orthogonal basis data 810.

[0049] The orthogonal basis data 810 is data of an orthogonal basis used in a transformation function that generates a time domain signal whose frequency spectrum changes over time by weighting a predetermined orthogonal basis with a weight. As such a transformation function, an inverse wavelet transform or the like can be used, and when the inverse wavelet transform is used, the orthogonal basis data 810 becomes data of a wavelet basis. Here, the wavelet basis is generally obtained by varying the scale and time shift of a mother wavelet in various ways.

[0050] The time-frequency analysis unit 807 performs a time-frequency analysis on the observation sound data output by the microphone 801, and converts the observation sound data into time-frequency domain observation sound data representing the weighting coefficients of the orthogonal basis of the above-mentioned conversion function. Here, when the orthogonal basis data 810 is wavelet basis data, the time-frequency analysis performed by the frequency analysis unit uses wavelet transform, and the time-frequency domain observed sound data becomes wavelet coefficients. The update control unit 806 stops the operation of the adaptive filter 803 and causes the ICA processing unit 805 to perform independent component analysis processing when setting up the noise control device or when it is detected from the time-frequency domain observation sound data that the sound level represented by the observation sound data output by the microphone 801 has increased to a predetermined level or more.

[0051] In the independent component analysis process, the ICA processing unit 805 performs independent component analysis in the time-frequency domain, separates data representing noise components in the time-frequency domain from the time-frequency domain observed sound data based on statistical independence, and sends the separated data to the basis spectrum calculation unit 809 as time-frequency domain statistical noise data.

[0052] The basis spectrum calculation unit 809 calculates the spectrum for each element of the orthogonal basis stored in the orthogonal basis data 810, of the noise represented in the time-frequency domain by the time-frequency domain statistical noise data sent from the ICA processing unit 805, and sets the weighting coefficient of the orthogonal basis obtained from the calculated spectrum as the weighting coefficient 804.

[0053] Furthermore, when a wavelet transform is used as the frequency analysis performed by the frequency analysis unit, the weighting coefficients stored in weighting coefficient 804 are wavelet coefficients. Once the weighting coefficients 804 have been stored, the update control unit 806 resumes the operation of the adaptive filter 803 . The stored weighting coefficients 804 are output to the adaptive filter 803 as time-frequency domain pseudo noise data representing the weighting coefficients of the orthogonal basis stored in the orthogonal basis data 810 . The adaptive filter 803 performs filtering on the input time-frequency domain artificial noise data in the frequency domain and outputs the result as time-frequency domain noise-canceled sound data to the time domain conversion unit 808. The adaptive filter 803 also treats the time-frequency domain observed sound data output by the time-frequency analysis unit 807 as an error and adapts the filter coefficients of the filtering performed on the time-frequency domain artificial noise data using an adaptive algorithm such as LMS or NLMS so as to minimize the power of the error.

[0054] The time domain transform section 808 transforms the time-frequency domain noise-canceled sound data output by the adaptive filter 803 into sound in the time domain, and outputs the sound from the speaker 802 as the noise-canceled sound NC. Here, when wavelet transform is used as the frequency analysis performed by the time-frequency analysis unit 807, the transform performed by the time domain transform unit 808 is an inverse wavelet transform. In the above-described eighth embodiment, instead of the ICA processing unit 15, a processing unit may be provided that separates data representing noise in the time-frequency domain from the time-frequency domain observed sound data by using a trained neural network or principal component analysis, and converts the separated data into time-frequency domain statistical noise data.

[0055] The eighth embodiment of the present invention has been described above. According to each of the above embodiments, noise components are analyzed and stored from a sound picked up by a microphone that picks up errors for adapting an adaptive filter, and pseudo noise data representing the stored noise components is used as an input to an adaptive filter to generate a noise-canceled sound by active noise control. Furthermore, the problem of feedback loop propagation delay, which is a cause of noise not being able to be cancelled well in the high frequency range as in feedback-type active noise control, does not occur. Furthermore, since pre-stored pseudo noise data is used as an input to the adaptive filter, the problem of insufficient calculation processing time, which is a cause of noise not being able to be cancelled well in the high frequency range in feedforward-side active noise control when a sensor that detects an event correlated with noise at a position upstream of the noise cancellation target position is close to the noise cancellation target position, does not occur. Therefore, according to each embodiment, in a blind type configuration in which a sensor for detecting an event correlated with noise is not provided at a position upstream of the noise cancellation target position, active noise control can be performed that can effectively cancel noise even in the high frequency range.

[0056] Here, as the weighting coefficient 704 in the seventh embodiment described above, a weighting coefficient that is fixedly set in advance may be used. That is, in this case, as shown in FIG. 9a, the ICA processing unit 705, the update control unit 706, and the basis spectrum calculation unit 709 are eliminated, and the weighting coefficients of the orthogonal basis determined by a prior analysis or the like are set in advance as the weighting coefficients 704. Similarly, a weighting coefficient that is fixedly set in advance may be used as the weighting coefficient 804 in the eighth embodiment. That is, in this case, as shown in FIG. 9b, the ICA processing unit 805, the update control unit 806, and the basis spectrum calculation unit 809 are eliminated, and the weighting coefficients of the orthogonal basis determined by a prior analysis or the like are set in advance as the weighting coefficients 804. With these configurations, noise can also be effectively cancelled in the high frequency range in a blind type configuration in which no sensor is provided for detecting an event correlated with noise at a position upstream of the noise cancellation target position. [Explanation of symbols]

[0057] 11...microphone, 12...speaker, 13...adaptive filter, 14...statistical noise data, 15...ICA processing unit, 16...update control unit, 21...microphone, 22...speaker, 23...adaptive filter, 24...frequency domain statistical noise data, 25...ICA processing unit, 26...update control unit, 27...frequency analysis unit, 28...time domain conversion unit, 31...microphone, 33...adaptive filter, 34...inferred noise data, 35...trained neural network, 36...update control unit, 41...microphone, 42...speaker, 43...adaptive filter, 44...frequency domain inferred noise data, 45...trained neural network, 46...update control unit, 47...frequency analysis unit, 48...time domain conversion unit, 51...microphone, 52...speaker, 53...adaptive filter, 54...principal component data, 55...principal component analysis unit, 56...update control unit, 61...microphone, 62...speaker, 63...adaptive filter, 64...frequency domain principal component data, 65...principal component analysis unit, 66...update control unit, 67...frequency analysis unit, 68...time domain transformation unit, 701...microphone, 702...speaker, 703...adaptive filter, 704...weighting coefficient, 705...ICA processing unit, 706...update control unit, 707...frequency analysis unit, 708...time domain transformation unit, 709...basis spectrum calculation unit, 710...orthogonal basis data, 801...microphone, 802...speaker, 803...adaptive filter, 804...weighting coefficient, 805...ICA processing unit, 806...update control unit, 807...time-frequency analysis unit, 808...time domain transformation unit, 809...basis spectrum calculation unit, 810...orthogonal basis data.

Claims

1. An active noise control system that performs active noise control, A microphone that outputs the picked up sound as an observation signal; a speaker that outputs a noise canceling sound; An adaptive filter; A storage means; a noise separation means for performing a noise separation process at a predetermined timing, the noise separation process being for separating a noise component from the observed signal and storing a pseudo-noise signal representing the separated noise component in the storage means; an adaptive filter to which the pseudo noise signal stored in the memory means is input, The active noise control system is characterized in that the adaptive filter performs an adaptive operation using the observed signal as an error and generates a noise-canceling sound to be output from the speaker from the artificial noise signal.

2. An active noise control system that performs active noise control, With a microphone a frequency analysis means for converting the output of the microphone into a signal in a frequency domain and outputting the signal as an observation signal; a speaker that outputs a noise canceling sound; An adaptive filter; A storage means; time domain transformation means; a noise separation means for performing a noise separation process at a predetermined timing, the noise separation process being for separating a noise component from the observed signal in a frequency domain and storing a pseudo-noise signal representing the separated noise component in the frequency domain in the storage means; an adaptive filter to which the pseudo noise signal stored in the memory means is input, The adaptive filter performs an adaptive operation in a frequency domain using the observed signal as an error; The active noise control system according to claim 1, wherein the time domain conversion means converts the output of the adaptive filter into a time domain signal to generate a noise-canceling sound to be output from the speaker.

3. 3. An active noise control system according to claim 2, The frequency analysis means converts the output of the microphone into a frequency domain signal by fast Fourier transform and outputs the signal as an observation signal; 2. An active noise control system according to claim 1, wherein said time domain transform means transforms the output of said adaptive filter into a time domain signal by inverse fast Fourier transform.

4. An active noise control system that performs active noise control, With a microphone an analysis means for converting the output of the microphone into a weighting coefficient of a predetermined conversion function that generates a time domain signal by weighting an orthogonal basis, and outputting the converted signal as an observation signal; a speaker that outputs a noise canceling sound; An adaptive filter; A storage means; time domain transformation means; a noise separation means for performing a noise separation process at a predetermined timing, which separates a noise component from the observed signal and stores the noise component as a pseudo-noise signal in the storage means; an adaptive filter to which the pseudo noise signal stored in the memory means is input, The adaptive filter performs an adaptive operation using the observed signal as an error; The active noise control system according to claim 1, wherein the time domain transformation means transforms the output of the adaptive filter into a time domain signal using the transformation function, and generates a noise-canceling sound to be output from the speaker.

5. 5. An active noise control system according to claim 4, The analysis means converts the microphone output into the weighting coefficients by a fast Fourier transform and outputs the weighting coefficients as an observation signal; 2. An active noise control system according to claim 1, wherein said time domain transform means transforms the output of said adaptive filter into a time domain signal by inverse fast Fourier transform.

6. 5. An active noise control system according to claim 4, The analysis means converts the microphone output into the weighting coefficients by wavelet transformation and outputs the weighting coefficients as an observation signal; 2. An active noise control system, comprising: a time domain transforming means for transforming an output of the adaptive filter into a time domain signal by inverse wavelet transform.

7. 7. An active noise control system according to claim 1, 2, 3, 4, 5 or 6, comprising: An active noise control system, characterized in that the noise separation means performs the noise separation process when the output of the microphone becomes larger than a predetermined level.

8. 7. An active noise control system according to claim 1, 2, 3, 4, 5 or 6, comprising: An active noise control system, wherein the noise separation means separates noise components from the observed signal by independent component analysis (ICA).

9. 7. An active noise control system according to claim 1, 2, 3, 4, 5 or 6, comprising: An active noise control system, wherein the noise separating means infers and separates the noise components from the observation signals using a neural network that has been trained in advance to infer the noise components from the observation signals.

10. 7. An active noise control system according to claim 1, 2, 3, 4, 5 or 6, comprising: An active noise control system, wherein the noise separation means calculates principal components of a predetermined order of the observed signal by principal component analysis, and separates the calculated principal components as the noise components.

11. An active noise control system that performs active noise control, With a microphone an analysis means for converting the output of the microphone into a weighting coefficient of a predetermined conversion function that generates a time domain signal by weighting an orthogonal basis, and outputting the converted signal as an observation signal; a speaker that outputs a noise canceling sound; An adaptive filter; a storage means for storing in advance, as a pseudo noise signal, the weighting coefficient for generating a pseudo noise by the conversion function; time domain transformation means; an adaptive filter to which the pseudo noise signal stored in the memory means is input, The adaptive filter performs an adaptive operation using the observed signal as an error; The active noise control system according to claim 1, wherein the time domain transformation means transforms the output of the adaptive filter into a time domain signal using the transformation function, and generates a noise-canceling sound to be output from the speaker.

12. 12. An active noise control system according to claim 11, comprising: The analysis means converts the microphone output into the weighting coefficients by a fast Fourier transform and outputs the weighting coefficients as an observation signal; 2. An active noise control system according to claim 1, wherein said time domain transform means transforms the output of said adaptive filter into a time domain signal by inverse fast Fourier transform.

13. 12. An active noise control system according to claim 11, comprising: The analysis means converts the microphone output into the weighting coefficients by wavelet transformation and outputs the weighting coefficients as an observation signal; 2. An active noise control system, comprising: a time domain transforming means for transforming an output of the adaptive filter into a time domain signal by inverse wavelet transform.

Images