JP4725422B2 - Echo cancellation circuit, acoustic device, network camera, and echo cancellation method - Google Patents

Description

  The present invention relates to an echo cancellation circuit that reduces echoes, an audio device using the echo cancellation circuit, and a network camera. The present invention also relates to an echo cancellation method used in such an echo cancellation circuit.

  An audio device including a speaker and a microphone, in particular, a voice interactive communication device capable of performing bidirectional audio communication with others, such as a telephone, a mobile phone terminal device, a video conference system, and a network camera Various audio two-way communication devices are used. When such a voice bidirectional communication device is used, a speaker of the other voice bidirectional communication device installed at a remote location, for example, according to the voice signal transmitted from the voice interactive communication device on one speaker side Audio output. When the voice output from the speaker is picked up by the microphone of the other voice bidirectional communication device, the voice is transmitted as an echo to the voice two-way communication device on the speaker side and becomes noise. Conversations are disturbed.

Therefore, as a means for reducing such echoes, a so-called acoustic echo cancellation method is known (see, for example, Patent Document 1). In the acoustic echo cancellation method, first, a coefficient of an impulse response function in the environment, that is, a so-called impulse response coefficient is estimated from an input signal acquired by a microphone when sound is output from a speaker. Then, using the estimated impulse response coefficient, an echo component is estimated from the audio output signal to the speaker, and this echo component is subtracted from the microphone input signal, thereby causing the microphone to pick up the audio output of the speaker. Reduce noise of echo component.
Japanese Patent Laid-Open No. 7-66757

  By the way, in the acoustic echo cancellation method as described above, the least square method or the like is used to estimate the impulse response coefficient. In the least square method, while changing the impulse response coefficient h, the sum of squares of the estimation error r, which is the difference between the echo component estimated using the impulse response coefficient h and the actual echo component acquired by the microphone, is repeated. The impulse response coefficient h is calculated so that the sum of squares of the estimation error r is minimized. As described above, in order to calculate the estimated value of the impulse response coefficient h by the least square method, the calculation for calculating the sum of squares of the estimation error r while changing the impulse response coefficient h is generally performed several thousand times. It needs to be repeated about tens of thousands of times, and it takes about several seconds to calculate the estimated value of the impulse response coefficient h. Then, in an acoustic device such as a voice two-way communication device using such an acoustic echo cancellation method, it is not possible to reduce the noise generated by the echo component for the time until the estimated value of the impulse response coefficient h is calculated. There was an inconvenience.

  The present invention has been made in view of such circumstances, and provides an echo cancellation circuit, an acoustic device, a network camera, and an echo cancellation method that can quickly reduce noise caused by an echo component. With the goal.

An echo cancellation circuit according to the present invention includes an audio output signal receiving unit that receives a signal representing sound as an audio output signal, an audio output terminal that outputs an audio output signal received by the audio output signal receiving unit, A voice input terminal for receiving a voice input signal, an impulse response coefficient representing a correlation between the voice output signal and the voice input signal, an initial value storage unit for storing in advance as an initial impulse response coefficient, and the initial value A new impulse response based on the audio output signal received by the audio output signal receiving unit and the audio input signal received by the audio input terminal, with the initial impulse response coefficient stored in the storage unit as an initial value. An impulse response coefficient calculating unit for calculating a coefficient; an audio output signal received by the audio output signal receiving unit; and the impulse response unit From an impulse response coefficient calculated by the calculation unit, an estimated echo signal generation unit that estimates an audio input signal received by the audio input terminal as an estimated echo signal, and an audio input signal received by the audio input terminal A synthesized signal output unit that outputs a synthesized signal obtained by subtracting the estimated echo signal generated by the estimated echo signal generating unit , and the initial value storage unit that stores the new impulse response coefficient. A limit value storage unit that preliminarily stores an estimated range limit value indicating an allowable variation range with respect to the initial impulse response coefficient, and the impulse response coefficient calculated by the impulse response coefficient calculation unit is the limit value storage unit. And an estimation range limiter that adjusts to limit the range within the range of the estimation range limit value stored.

  According to this configuration, the audio output signal representing the sound is received by the audio output signal receiving unit, and the audio output signal received by the audio output signal receiving unit is output by the audio output terminal. The initial value storage unit stores in advance an impulse response coefficient representing a correlation between the audio output signal output from the audio output terminal and the audio input signal received by the audio input terminal as an initial impulse response coefficient. Then, the impulse response coefficient calculation unit sets the initial impulse response coefficient stored in the initial value storage unit as an initial value, and the voice output signal received by the voice output signal reception unit and the voice input signal received by the voice input terminal Based on the above, a new impulse response coefficient is calculated. Further, the estimated echo signal generation unit estimates the voice input signal received by the voice input terminal from the voice output signal received by the voice output signal reception unit and the impulse response coefficient calculated by the impulse response coefficient calculation unit. Estimated as an echo signal. Then, the synthesized signal output unit outputs a synthesized signal obtained by subtracting the estimated echo signal generated by the estimated echo signal generating unit from the audio input signal actually received by the audio input terminal.

  According to this, when a new impulse response coefficient is calculated by the impulse response coefficient calculation unit, the initial impulse response coefficient stored in the initial value storage unit is used as the initial value of the arithmetic processing. The coefficient calculation processing time is shortened. Then, the estimated echo signal generation unit can quickly start the estimated echo signal generation process using the impulse response coefficient and quickly generate the estimated echo signal. The synthesized signal output unit can output a synthesized signal in which noise caused by the echo component is reduced by subtracting the estimated echo signal from the voice input signal received by the voice input terminal. Can be quickly reduced.

  According to this configuration, the estimated value limit value indicating the variation range allowed for the initial impulse response coefficient stored in the initial value storage unit is stored in advance by the limit value storage unit. Then, the impulse response coefficient calculated by the impulse response coefficient calculation unit is limited by the estimation range limit unit within the range of the estimation range limit value stored in the limit value storage unit, so that the impulse response coefficient is allowed. It is possible to suppress a situation in which noise greatly varies beyond the fluctuation range, and noise caused by echo components cannot be reduced due to a long calculation time.

  The initial value storage unit receives an initial setting instruction for storing an impulse response coefficient. When the initial setting instruction is received by the receiving unit, the initial value storage unit calculates the impulse response coefficient. It is preferable to further include an initial setting control unit that stores the new impulse response coefficient as the initial impulse response coefficient in the initial value storage unit.

  According to this configuration, when the initial setting instruction is received by the receiving unit, the new impulse response coefficient calculated by the impulse response coefficient calculating unit is stored in the initial value storage unit as the initial impulse response coefficient. After the echo cancellation circuit, the audio output unit, and the audio input unit are installed, an initial setting instruction is received by the receiving unit, so that an initial impulse response coefficient corresponding to the installation environment can be stored in the initial value storage unit.

  An acoustic device according to the present invention includes the above-described echo cancellation circuit, a sound output unit that converts a sound output signal output from the sound output terminal into a sound wave, and a sound wave that is converted into the sound input signal. A housing that accommodates an audio input unit that outputs to an input terminal, the echo cancellation circuit, the audio output unit, and the audio input unit so that the positional relationship between the audio output unit and the audio input unit is fixed. And the impulse response coefficient is a voice input received from the voice input unit by the voice input terminal when the voice output signal output from the voice output terminal is converted into a sound wave by the voice output unit. The correlation between the signal and the audio output signal is shown.

  According to this configuration, since the positional relationship between the audio output unit and the audio input unit is fixed by the housing, the actual use environment with respect to the initial impulse response coefficient stored in advance in the initial value storage unit As a result, the impulse response coefficient calculation processing time is shortened. Then, the estimated echo signal generation unit can quickly start the estimated echo signal generation process using the impulse response coefficient and quickly generate the estimated echo signal. The synthesized signal output unit can output a synthesized signal in which noise caused by the echo component is reduced by subtracting the estimated echo signal from the voice input signal received by the voice input terminal. Can be quickly reduced.

  A network camera according to the present invention includes an echo cancellation circuit according to any one of the above, a voice output unit that converts a voice output signal output from the voice output terminal into a sound wave, and a sound wave as the voice input signal. An audio input unit that converts the image data into an audio input terminal, an image capturing unit that captures an image to acquire image data, and a terminal device that connects the image data acquired by the image capturing unit via a network The audio output signal receiving unit receives the audio output signal from the terminal device via the network, and the combined signal output unit is configured to transmit the audio output signal via the network. The combined signal is transmitted to the terminal device.

  According to this configuration, the echo cancel circuit, the sound output unit, the sound input unit, the image capturing unit that captures an image to acquire image data, and the image data acquired by the image capturing unit are connected to the network. In a network camera including an image data output unit that transmits to a terminal device connected via the network, noise caused by echo components can be quickly reduced.

The echo canceling method according to the present invention includes a sound output signal receiving step for receiving a signal representing sound as a sound output signal, and a sound output for converting the sound output signal received by the sound output signal receiving step into a sound wave. A voice input step for converting a sound wave into a voice input signal, and a voice input step when the voice output signal received by the voice output signal receiving step is converted into a sound wave in the voice output step. An initial value storage step of storing in the initial value storage unit in advance an impulse response coefficient representing a correlation between the audio input signal and the audio output signal, and an initial value stored in the initial value storage unit With the impulse response coefficient as an initial value, the voice output signal received by the voice output signal receiving step and the voice input step are obtained. An impulse response coefficient calculating step for calculating a new impulse response coefficient based on the voice input signal, an audio output signal received by the voice output signal receiving step, and an impulse response coefficient calculated by the impulse response coefficient calculating step From the above, an estimated echo signal estimation step for generating an audio input signal obtained by the audio input step based on the sound wave output in the audio output step as an estimated echo signal, and an audio input signal obtained by the audio input step From the estimated echo signal generated by the estimated echo signal estimation step, a synthesized signal output step of outputting a synthesized signal obtained by subtracting, and the new impulse response coefficient stored in the initial value storage unit Estimated range indicating the allowable variation range for the initial impulse response coefficient Limiting the limit value in the limit value storage unit and limiting the impulse response coefficient calculated in the impulse response coefficient calculation step within the range of the estimated range limit value stored in the limit value storage unit And an estimated range limiting step for adjusting in such a manner .

  According to this configuration, an audio output signal representing sound is received, and the audio output signal received by the audio output signal receiving unit is converted into sound waves. Further, an impulse response coefficient representing a correlation between the voice input signal obtained when the voice output signal is converted into a sound wave and the voice output signal is stored in advance in the initial value storage unit as an initial impulse response coefficient. Then, using the initial impulse response coefficient stored in the initial value storage unit as an initial value, a new impulse response coefficient is calculated based on the audio output signal and the audio input signal. Furthermore, the voice input signal obtained based on the sound wave is estimated as the estimated echo signal from the voice output signal and the impulse response coefficient. Then, a synthesized signal obtained by subtracting the estimated echo signal from the voice input signal is output.

  According to this, when a new impulse response coefficient is calculated, the initial impulse response coefficient stored in the initial value storage unit is used as the initial value of the calculation process, so that the calculation time of the impulse response coefficient is shortened. Is done. Then, the estimated echo signal can be generated quickly by accelerating the start of the process of generating the estimated echo signal using the impulse response coefficient. And since the synthetic | combination signal which reduced the noise produced by an echo component by subtracting an estimated echo signal from an audio | voice input signal can be output, the noise produced by an echo component can be reduced rapidly.

  According to this configuration, the estimated range limit value indicating the variation range allowed for the initial impulse response coefficient stored in the initial value storage unit is stored in advance. In the estimated range limiting step, the impulse response coefficient calculated in the impulse response coefficient calculating step is limited within the range of the estimated range limit value stored in the limit value storage unit, so that the impulse response coefficient is allowed. It can be suppressed that noise generated by echo components cannot be reduced due to large fluctuations exceeding the fluctuation range.

  In the echo cancellation circuit, the acoustic device, the network camera, and the echo cancellation method configured as described above, the initial impulse response coefficient stored in the initial value storage unit is calculated when a new impulse response coefficient is calculated. Since it is used as an initial value, the impulse response coefficient calculation processing time is shortened. Then, the estimated echo signal can be generated quickly by accelerating the start of the process of generating the estimated echo signal using the impulse response coefficient. And since the synthetic | combination signal which reduced the noise produced by an echo component by subtracting an estimated echo signal from an audio | voice input signal can be output, the noise produced by an echo component can be reduced rapidly.

  Embodiments according to the present invention will be described below with reference to the drawings. In addition, the structure which attached | subjected the same code | symbol in each figure shows that it is the same structure, The description is abbreviate | omitted. FIG. 1 is a perspective view showing an example of the appearance of a network camera, which is an example of an audio device using an echo cancellation circuit according to an embodiment of the present invention.

  In the network camera 1 shown in FIG. 1, a speaker 2 (audio output unit), a microphone 3 (audio input unit), and an imaging unit 4 are accommodated in a housing 5 formed by, for example, a substantially box-shaped resin molding. Configured. The speaker 2 is fixedly disposed on the upper left of the front surface of the housing 5, and the microphone 3 is disposed on the substantially diagonal line on the front surface of the housing 5. With such an arrangement, the distance between the speaker 2 and the microphone 3 is increased, and acquisition of sound output from the speaker 2 with the microphone 3 is reduced.

  FIG. 2 is a block diagram showing an example of the configuration of the network camera 1 shown in FIG. The network camera 1 shown in FIG. 2 includes a speaker 2, a microphone 3, an imaging unit 4, an image encoder 41, a communication I / F unit 61, and an echo cancellation circuit 6. 2 includes a connection terminal 612 (audio output signal receiving unit), a DA converter 614, an audio output terminal 21, an audio input terminal 31, an AD converter 615, a connection terminal 613 (synthetic signal output unit), and an initial stage. A value storage unit 62, a limit value storage unit 63, a push button switch 64 (accepting unit), and an arithmetic processing unit 65 are provided.

  The communication I / F unit 61 is connected to a terminal device such as a personal computer or another network camera 1 via a network 611 so as to be able to transmit and receive data. Then, the communication I / F unit 61 samples a signal transmitted from a terminal device connected to the network 611, for example, a signal representing sound, at a predetermined frequency, for example, 8 kHz (sampling period is 125 μsec). Then, the communication I / F unit 61 converts the sampling value into an audio output signal s in a format that can be converted into audio by the speaker 2, and outputs it to the speaker 2 and the arithmetic processing unit 65 via the connection terminal 612. In addition, the communication I / F unit 61 converts the composite signal r generated by the arithmetic processing unit 65 and the image data g generated by the image encoder 41 into a communication signal according to the communication protocol of the network 611, and the network 611. To other terminal devices. In this case, the communication I / F unit 61 corresponds to an example of an audio output signal receiving unit, a synthesized signal output unit, and an image data output unit.

  The connection terminals 612 and 613 are, for example, through holes and pads for soldering connectors and wiring. The image data output unit may be, for example, a connector for outputting the image data g to the outside, a through hole or a pad for soldering the wiring.

  The DA converter 614 converts the audio output signal s into an analog signal and outputs the analog signal to the speaker 2 via the audio output terminal 21.

  The audio output terminal 21 is a connection terminal that outputs the output signal of the DA converter 614 to the speaker 2, and may be, for example, a through hole or a pad for soldering a connector or wiring of the speaker 2. The audio input terminal 31 is connected to the microphone 3 and outputs the audio input signal output from the microphone 3 to the AD converter 615. The audio input terminal 31 may be, for example, a through hole or a pad for soldering a connector or a wiring of the microphone 3.

  The AD converter 615 converts the voice input signal output from the microphone 3 into a digital signal and outputs the digital signal to the adder 68 as the voice input signal e. The speaker 2 converts the sound output signal output from the sound output terminal 21 into a sound wave. The microphone 3 converts the sound wave into an audio input signal and outputs it to the audio input terminal 31.

  The initial value storage unit 62 and the limit value storage unit 63 are configured using, for example, an EEPROM (Electrically Erasable and Programmable Read Only Memory) which is a nonvolatile memory. The initial value storage unit 62 is connected to the audio input terminal 31 when the audio output signal s received by the communication I / F unit 61 is converted into a sound wave by the speaker 2 connected to the audio output terminal 21. An impulse response coefficient h representing the correlation between the audio input signal e obtained by the microphone 3 and the audio output signal s is stored in advance.

  Further, the limit value storage unit 63 permits the new impulse response coefficient h calculated by the arithmetic processing unit 65 with respect to the impulse response coefficient h (initial impulse response coefficient) stored in the initial value storage unit 62. An estimated range limit value indicating the fluctuation range to be used is stored in advance.

  The push button switch 64 is a push button switch that can be operated by the user. Then, when the user presses the push button switch 64, a signal indicating an initial setting instruction for storing the new impulse response coefficient h in the initial value storage unit 62 is sent from the push button switch 64 to the arithmetic processing unit 65. Is output. In this case, the push button switch 64 corresponds to an example of a reception unit.

  The reception unit is not limited to an operation switch such as the push button switch 64. The initial setting instruction may be transmitted from the terminal device connected to the network 611 to the communication I / F unit 61 as a communication command via the network 611, for example. In this case, the communication I / F unit 61 corresponds to an example of a reception unit.

  The arithmetic processing unit 65 is a signal processing circuit that generates a composite signal r obtained by removing an echo component from the audio input signal e based on the audio output signal s and the audio input signal e, and outputs the synthesized signal r to the communication I / F unit 61. is there. The arithmetic processing unit 65 is configured using, for example, a DSP (Digital Signal Processor), and functions as an impulse response estimation unit 66, an estimated echo signal generation unit 67, and an adder 68 by executing a predetermined control program. To do.

  The impulse response estimation unit 66 functions as an impulse response coefficient calculation unit 661, an estimation range restriction unit 662, an impulse response coefficient storage unit 663, and an initial setting control unit 664. The estimated echo signal generation unit 67 functions as a sampling value storage unit 672 and an echo signal estimation calculation unit 673. Note that the arithmetic processing unit 65 is not limited to an example configured using a DSP, and may be configured as a dedicated circuit using, for example, an ASIC (Application Specific Integrated Circuit).

  The impulse response coefficient calculation unit 661 uses the impulse response coefficient h stored in the initial value storage unit 62 as an initial value, and the audio output signal s obtained by the communication I / F unit 61 and the audio input terminal 31 from the microphone 3. A new impulse response coefficient h is calculated on the basis of the voice input signal e received by.

  The estimated range limiting unit 662 adjusts the impulse response coefficient h calculated by the impulse response coefficient calculating unit 661 so as to be limited within the range of the estimated range limit value stored in the limit value storage unit 63, and the impulse response The coefficient is stored in the coefficient storage unit 663. The impulse response coefficient storage unit 663 is a storage unit configured by, for example, a register circuit or a RAM (Random Access Memory).

  When the push button switch 64 is pressed, the initial setting control unit 664 stores the impulse response coefficient h stored in the impulse response coefficient storage unit 663 in the initial value storage unit 62 and an initial value of a new impulse response coefficient h. Remember as.

  Note that when the push button switch 64 is pressed, the initial setting control unit 664 does not use the impulse response coefficient h stored in the initial value storage unit 62 by the impulse response coefficient calculation unit 661 and does not use the communication I / F unit. 61, a new impulse response coefficient h is calculated based only on the audio output signal s obtained by 61 and the audio input signal e received by the audio input terminal 31 from the microphone 3, and this impulse response coefficient h is set as an initial value. It is good also as a structure memorize | stored in the value memory | storage part 62 as an initial value. Thereby, the impulse response coefficient h can be calculated without using the initial value of the impulse response coefficient h and stored in the initial value storage unit 62. Therefore, the initial value of the impulse response coefficient h is stored in the initial value storage unit 62. Even if not stored, the impulse response coefficient h can be calculated and the initial value storage unit 62 can store the initial value of the impulse response coefficient h.

  The sampling value storage unit 672 is a storage unit configured by, for example, a register circuit or a RAM. The sampling value storage unit 672 stores, for example, n values of the audio output signal s sampled by the communication I / F unit 61 as s [1] to s [n].

  In this case, s [n] indicates the current value of the most recently sampled audio output signal s, that is, the audio output signal s. For example, when the number n of sampling values of the audio output signal s stored in the sampling value storage unit 672 is 256, s [256] indicates the current value of the audio output signal s, and s [1] is s The past audio output signal s sampled 125 μsec × 255 = 32 msec before [256] is shown. In this case, “j” in s [j] indicates a sampling number. Similarly, in the following description, the voice input signal e [j], the estimated echo signal e ′ [j], and the synthesized signal r [j] are respectively the voice input signal e and the estimated echo signal e corresponding to the sampling number j. 'And the synthesized signal r. Further, the voice input signal e [n], the estimated echo signal e ′ [n], and the synthesized signal r [n] each indicate a current value.

  The echo signal estimation calculation unit 673 performs a convolution calculation from the sampling value of the audio output signal s stored in the sampling value storage unit 672 and the impulse response coefficient h stored in the impulse response coefficient storage unit 663. Thus, the estimated value of the voice input signal e obtained by the microphone 3 based on the sound wave output from the speaker 2, that is, the estimated value of the voice input signal e of the echo component is generated as the estimated echo signal e ′, and the estimated echo signal e The sign of 'is inverted and output to the adder 68.

  The adder 68 adds the audio input signal e and the estimated echo signal e ′ whose polarity is inverted, thereby subtracting the estimated echo signal e ′ from the audio input signal e to generate a synthesized signal r. The adder 68 outputs the combined signal r to the impulse response estimation unit 66 and also outputs it to the communication I / F unit 61 via the connection terminal 613.

  The imaging unit 4 is a camera that captures an image and obtains image data p, and includes an imaging element such as a complementary metal oxide semiconductor (CMOS) sensor or a charge coupled device (CCD) sensor, and a lens. Yes.

  The image encoder 41 decodes the image data p acquired by the imaging unit 4 and outputs it to the communication I / F unit 61 as image data g in a data format of MPEG (Moving Picture Experts Group) 4, for example.

  Next, the operation of the network camera 1 configured as described above will be described. The network camera 1 shown in FIG. 2 needs to store the initial value of the impulse response coefficient h in the initial value storage unit 62 in advance, but still stores the initial value when the network camera 1 is manufactured in a factory, for example. The impulse response coefficient h is not stored in the unit 62.

  Therefore, an operation for storing the initial value of the impulse response coefficient h in the initial value storage unit 62 will be described. First, the initial value storage unit 62 does not store the initial value of the impulse response coefficient h, and the initial value storage unit 62 is in an initial state. For example, “zero” is stored in the entire storage area. .

  Next, for example, when the network camera 1 is shipped from the factory, the operator may cause the network camera 1 to be in a silent environment with no external sound, and the sound output from the speaker 2 may be reflected by a wall or ceiling. It arrange | positions in the environment which does not reach the microphone 3, for example, an anechoic room. When the worker transmits the test voice data to the communication I / F unit 61 via the network 611 using the terminal device, the communication I / F unit 61 converts the voice output signal s into the impulse response estimation unit. 66, the estimated echo signal generator 67, and the speaker 2.

  Then, the sound output from the speaker 2 is acquired by the microphone 3 as an echo w, and the sound input signal e is output from the microphone 3. In this case, since the sound output from the speaker 2 is acquired by the microphone 3 via the housing of the network camera 1 and the surrounding air, based on the sound input signal e thus obtained, By calculating the impulse response coefficient h by the impulse response coefficient calculation unit 661, the impulse response coefficient h unique to the network camera 1 can be calculated.

  Further, the adder 68 subtracts the estimated echo signal e ′ from the audio input signal e, and outputs the synthesized signal r to the impulse response coefficient calculator 661. Then, the audio output signal s is sampled at 8 kHz, for example, by the communication I / F unit 61, and the sampling values s [1] to s [n] are stored in the sampling value storage unit 672.

  Next, based on the audio output signal s obtained by the communication I / F unit 61 and the audio input signal e received from the microphone 3 by the audio input terminal 31 by the impulse response coefficient calculating unit 661, the impulse response coefficient is calculated. h is calculated. Specifically, the impulse response coefficient calculation unit 661 uses the sampling values s [1] to s [n] stored in the sampling value storage unit 672 and the current value of the combined signal r output from the adder 68. Based on a certain r [n], impulse response coefficients h [1] to h [n] respectively corresponding to the sampling values s [1] to s [n] are calculated by the least square method.

  In this case, the calculation process of the least square method is expressed by, for example, the following expression (1).

  In Expression (1), “←” indicates that the calculation result on the right side is assigned to the left side, and “j” indicates the sampling number.

  FIG. 3 is a flowchart showing an example of an operation for storing the initial value of the impulse response coefficient h in the initial value storage unit 62. First, the impulse response coefficient h is calculated by the pulse response coefficient calculation unit 661 (step S1). Since the sampling number is n and the sampling numbers are 1 to n, the impulse response coefficient calculation unit 661 calculates the impulse response coefficients h [1] to h [n] based on the above equation (1). Calculated.

  Here, when the right side of Expression (1) is first calculated, the impulse response coefficient h [j] stored in the initial value storage unit 62 is initially set as h [j] on the right side of Expression (1). Used as a value. However, the impulse response coefficient h [j] is not yet stored in the initial value storage unit 62, and “zero” is stored in the entire storage area. Then, the impulse response coefficient calculation unit 661 uses “zero” as the initial value of the impulse response coefficient h [j] to calculate the impulse response coefficients h [1] to h [n], and the impulse response coefficient storage unit 663.

  Next, the echo signal estimation calculation unit 673 stores the sampling values s [1] to s [n] of the audio output signal s stored in the sampling value storage unit 672 and the impulse response coefficient storage unit 663. By executing a convolution operation from the impulse response coefficients h [1] to h [n], an estimated echo signal e ′ [n] is generated. Then, the echo signal estimation calculation unit 673 outputs an estimated echo signal e ′ [n] whose polarity is inverted to the adder 68. Specifically, the echo signal estimation calculation unit 673 generates the estimated echo signal e ′ [n] based on, for example, the following equation (2).

  Then, the adder 68 generates a composite signal r [n] based on the following equation (3). In this case, the combined signal r [n] represents an estimation error r in the least square method.

r [n] = e [n] −e ′ [n] (3)
Then, the impulse response coefficient calculation unit 661, residue energy sigma _r of the estimation error r is or smaller than a preset predetermined threshold value R is checked as convergence criterion value of the least squares method (step S2), and If the residual energy σ _r exceeds the threshold value R (NO in step S2), the calculation process by the least square method has not converged, so the processes of steps S1 to S2 are repeated again. Here, in step S1, impulse response coefficients h [1] to h [n] stored in the impulse response coefficient storage unit 663 are used as h [j] on the right side of Expression (1). In this case, since it is necessary to converge the arithmetic processing by the least square method from the state of the initial value “zero”, the processing of steps S1 to S2 is repeated, for example, several thousand times to several tens of thousands times.

On the other hand, if the residual energy σ _r is equal to or less than the threshold value R (YES in step S2), the calculation process by the least square method has converged, and is stored in the impulse response coefficient storage unit 663 by the impulse response coefficient calculation unit 661. The impulse response coefficients h [1] to h [n] are stored in the initial value storage unit 62 as initial values (step S3).

  As described above, the initial values of the impulse response coefficients h [1] to h [n] are stored in the initial value storage unit 62 by the processes of steps S1 to S3. FIG. 4 is an explanatory diagram showing an example of the impulse response coefficient h thus obtained. In FIG. 4, the horizontal axis indicates the sampling number, and the vertical axis indicates the value of the impulse response coefficient h. In FIG. 4, the sampling number is equivalent to time t.

  For example, when the network camera 1 is mass-produced, it is necessary to store the initial values of the impulse response coefficients h [1] to h [n] in the initial value storage unit 62 by the processing of steps S1 to S3. No, the initial values of the impulse response coefficients h [1] to h [n] obtained by one network camera 1 may be stored as they are in the initial value storage unit 62 in the other network camera 1. .

  Next, an echo removal operation by the network camera 1 shown in FIG. 2 will be described. First, the network camera 1 is installed in an actual use environment such as an office or home. Then, the impulse response coefficient h (h [1] to h [n]) is acquired by the microphone 3 as the sound output signal s (s [1] to s [n]) and the echo w based on the sound output signal s. Therefore, it varies depending on the distance d between the speaker 2 and the microphone 3 and the environment where the network camera 1 is installed. For example, when the sound output from the speaker 2 is reflected by the wall of the room and acquired by the microphone 3, an echo appears in the audio input signal e.

  For this reason, if impulse response coefficients h [1] to h [n] are used as initial values in different installation environments, echo components depending on the environment, such as echoes reflected by the walls of the room, cannot be removed. However, the largest factor that affects the impulse response coefficients h [1] to h [n] is the distance d between the speaker 2 and the microphone 3. Then, since the positional relationship between the speaker 2 and the microphone 3 is fixed in the network camera 1, the distance d does not change, and the impulse response coefficients h [1] to h [n] even if the installation environment changes. Changes are reduced.

  FIG. 5 is an explanatory diagram for explaining an echo removal operation of the network camera 1. First, when a user transmits voice data to the communication I / F unit 61 via the network 611 using the terminal device, the voice output signal s shown in FIG. It is output to the estimation unit 66, the estimated echo signal generation unit 67, and the speaker 2. Then, the sound output from the speaker 2 is acquired by the microphone 3 as an echo w, and the sound input signal e shown in FIG.

  The sampling value storage unit 672 stores sampling values s [1] to s [n] of the audio output signal s sampled at 8 kHz, for example.

  Next, r, which is the current value of the sampling values s [1] to s [n] stored in the sampling value storage unit 672 and the combined signal r output from the adder 68 by the impulse response coefficient calculation unit 661. Based on [n], impulse response coefficients h [1] to h [n] adapted to the actual installation environment of the network camera 1 are calculated by processing similar to steps S1 to S2 shown in FIG. In this case, the initial value storage unit 62 stores initial values of impulse response coefficients h [1] to h [n] acquired in the silent environment, and is stored in the initial value storage unit 62 in step S1. Since the repetition calculation of steps S1 and S2 is performed using the impulse response coefficients h [1] to h [n] as initial values, the number of repetitions until the calculation process by the least square method converges is reduced. As a result, the calculation time of the impulse response coefficients h [1] to h [n] adapted to the installation environment of the network camera 1 is shortened.

  The network camera 1 is in a state where the microphone 3 is picking up sound other than the sound output from the speaker 2 such as a state in which the user is talking to the microphone 3 from the signal level of the audio input signal e, for example. A double-talk detection circuit (not shown) for detecting a double-talk state is provided. Then, when the double talk is not detected by the unillustrated double talk detection circuit, the impulse response coefficients h [1] to h [n] are calculated by the impulse response coefficient calculator 661. .

  Next, the impulse response coefficients h [1] to h [n] calculated by the impulse response coefficient calculating unit 661 by the estimated range limiting unit 662 are the ranges of the estimated range limit values stored in the limit value storage unit 63. Limited within. FIG. 6 is an explanatory diagram for explaining the operation of the estimation range restriction unit 662. In FIG. 6, impulse response coefficients h [1] to h [n] indicate impulse response coefficients h [1] to h [n] stored in the initial value storage unit 62. In addition, the estimated range limit values stored in the limit value storage unit 63 are indicated by A [1] to A [n]. The estimated range limit values A [1] to A [n] are stored in the limit value storage unit 63 corresponding to the impulse response coefficients h [1] to h [n], respectively. The estimated range limit values A [1] to A [n] are obtained by experimentally obtaining, for example, an allowable variation range for the impulse response coefficient h stored in the initial value storage unit 62, and limiting this in advance. What is necessary is just to make it memorize | store in the value memory | storage part 63. FIG.

  Note that the estimated range limit values A [1] to A [n] are not limited to those that set the range in a fixed manner. For example, the estimated range limit values A [1] to A [n] are those that set a variation ratio with respect to the impulse response coefficients h [1] to h [n]. There may be. Further, the estimated range limit values A [1] to A [n] are not limited to the examples set for the impulse response coefficients h [1] to h [n], for example, the impulse response coefficients h [1] to h [1] to The same estimated range limit value A may be set for h [n].

  Then, the impulse response coefficients h [1] to h [n] calculated by the impulse response coefficient calculation unit 661 by the estimation range restriction unit 662 are estimated range restriction values A [1] to A [n] shown in FIG. The impulse response coefficients h [1] to h [n] calculated by the impulse response coefficient calculation unit 661 are within the estimated range limit values A [1] to A [n]. If it is within the range, the impulse response coefficients h [1] to h [n] are stored in the impulse response coefficient storage unit 663, and the impulse response coefficients h [1] to h [n calculated by the impulse response coefficient calculation unit 661 are stored. ] Exceeds the range of the estimated range limit values A [1] to A [n], the upper limit value of the estimated range limit values A [1] to A [n] is the impulse response coefficient h [1] to h [ n] as an impulse response coefficient storage unit 66 If the impulse response coefficients h [1] to h [n] calculated by the impulse response coefficient calculation unit 661 are not less than the estimated range limit values A [1] to A [n], Lower limit values of values A [1] to A [n] are stored in the impulse response coefficient storage unit 663 as impulse response coefficients h [1] to h [n].

  Next, the echo signal estimation calculation unit 673 stores the sampling values s [1] to s [n] of the audio output signal s stored in the sampling value storage unit 672 and the impulse response coefficient storage unit 663. From the impulse response coefficients h [1] to h [n], an estimated echo signal e ′ [n] shown in FIG. 5C is generated based on, for example, Expression (2). Then, the echo signal estimation calculation unit 673 outputs an estimated echo signal e ′ [n] whose polarity is inverted to the adder 68. Further, the adder 68 generates a composite signal r [n] shown in FIG. 5D based on the equation (3). Then, the synthesized signal r [n] becomes a desired signal from which the echo component is removed.

  Further, the composite signal r [n] output from the adder 68 is transmitted to another terminal device via the network 611 by the communication I / F unit 61, thereby enabling bidirectional audio communication.

  As described above, according to the network camera 1, in the calculation process of the impulse response coefficients h [1] to h [n] by the impulse response coefficient calculation unit 661, the number of repetitions of steps S1 to S2 is reduced, and the network camera. Since the calculation time of impulse response coefficients h [1] to h [n] adapted to the installation environment of 1 is shortened, an echo cancel circuit 6 that can quickly reduce noise caused by an echo component, an example of an acoustic device The echo cancellation method used for the network camera 1 and the echo cancellation circuit 6 can be provided.

  Further, for example, the impulse response coefficients h [1] to h [n] are calculated by the impulse response coefficient calculation unit 661 by the above-described double talk detection circuit without detecting the double talk in spite of the double talk state. If the impulse response coefficients h [1] to h [n] are calculated under noisy conditions, the values of the impulse response coefficients h [1] to h [n] are greatly different from the correct values. In addition, based on the impulse response coefficients h [1] to h [n], the echo signal estimation calculation unit 673 generates the estimated echo signal e ′ [n] and the adder 68 generates the synthesized signal r [n]. Once generated, the echo component cannot be removed correctly. Further, if the values of the impulse response coefficients h [1] to h [n] are greatly different from the correct values, the calculation result of the least square method may not converge in the repeated calculation of steps S1 to S2.

  However, according to the network camera 1 shown in FIG. 2, the estimated range limiting unit 662 causes the fluctuation range of the impulse response coefficients h [1] to h [n] to be stored in the initial value storage unit 62. Since h [1] to h [n] are limited within the range of the estimated range limit values A [1] to A [n] stored in the limit value storage unit 63, for example, noise may occur depending on the surrounding environment. The impulse response coefficients h [1] to h [n] are greatly different from the appropriate values even in an environment such as outdoors where there is a large amount of noise, and the echo cannot be reduced. Accordingly, even in an environment where a correct impulse response coefficient h cannot be obtained due to noise or the like, a certain degree of echo cancellation performance can be obtained, and robustness to the environment is improved.

  By the way, as described above, it is obtained in an environment such as an anechoic room where there is no external sound and the sound output from the speaker 2 is reflected by a wall or ceiling and does not reach the microphone 3. When the impulse response coefficients h [1] to h [n] thus obtained are stored in the initial value storage unit 62 as initial values, this initial value is used in order to remove reflected sounds from walls and ceilings in the actual use environment. It is necessary to calculate new impulse response coefficients h [1] to h [n] by performing the least-squares iterative calculation in steps S1 and S2 using the values. In this case, the smaller the difference between the initial value and the convergence value of the iterative computation, the smaller the number of iterations of steps S1 to S2, and the calculation time of the impulse response coefficients h [1] to h [n] is shortened. Can do.

  Therefore, if the network camera 1 is not moved frequently, the impulse response coefficients h [1] to h [n] obtained in the installation environment of the network camera 1 can be stored in the initial value storage unit 62 as initial values. For example, in the subsequent calculation processing of the impulse response coefficients h [1] to h [n], the difference between the initial value and the convergence value of the iterative calculation decreases, and the impulse response coefficients h [1] to h [n] Calculation time can be shortened.

  Therefore, when the user presses the push button switch 64 after the network camera 1 is installed in the actual use environment, the impulse response coefficient h [1] stored in the impulse response coefficient storage unit 663 is stored by the initial setting control unit 664. ˜h [n] are stored in the initial value storage unit 62 as initial values of new impulse response coefficients h [1] to h [n].

  Note that the positional relationship between the speaker 2 and the microphone 3 does not necessarily have to be fixed. For example, unlike the audio apparatus 1a for the video conference system shown in FIG. 7, the speaker 2 is not fixed to the audio apparatus 1a, and the user freely arranges the speaker 2 in the actual use environment, and the audio output terminal 21 The speaker 2 may be connected to the wiring 23 for use. Further, for example, like the audio device 1b for the video conference system shown in FIG. 8, the microphone 3 is not fixed to the audio device 1b, and the user freely arranges the microphone 3 in the actual use environment, and the audio input terminal 31 and the microphone 3 may be used by being connected by the wiring 32. Alternatively, both the speaker 2 and the microphone 3 may be freely arranged.

  For example, when the positional relationship between the speaker 2 and the microphone 3 is not fixed as in the acoustic devices 1a and 1b described above, the distance d between the speaker 2 and the microphone 3 is large depending on the installation status of the speaker 2 and the microphone 3. Different. Then, the initial value of the impulse response coefficient h stored in advance in the initial value storage unit 62 at the time of factory shipment or the like is not necessarily an appropriate value. In such a case, the user installs the sound device such as the sound devices 1a and 1b in the actual use environment, installs the speaker 2 and the microphone 3, and then presses the push button switch 64, so that the installation environment is obtained. The corresponding impulse response coefficient h can be stored in the initial value storage unit 62 as an initial value.

  As an example of the audio device, a network camera and a video conference system are shown. However, the audio device only needs to use a microphone and a speaker, such as a mobile phone terminal device, a fixed phone, a karaoke device, and the like. May be.

It is a perspective view which shows an example of the external appearance of the network camera which is an example of the audio equipment using the echo cancellation circuit which concerns on one Embodiment of this invention. It is a block diagram which shows an example of a structure of the network camera shown in FIG. It is a flowchart which shows an example of the operation | movement which memorize | stores the initial value of an impulse response coefficient in an initial value memory | storage part. It is explanatory drawing which shows an example of an impulse response coefficient. It is explanatory drawing for demonstrating the echo removal operation | movement of the network camera shown in FIG. It is explanatory drawing for demonstrating operation | movement of the estimation range restriction | limiting part shown in FIG. It is a block diagram which shows an example of a structure of the audio equipment for video conference systems. It is a block diagram which shows another example of a structure of the audio equipment for video conference systems.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Network camera 1a, 1b Sound apparatus 2 Speaker 3 Microphone 4 Imaging part 5 Case 6 Echo cancellation circuit 21 Audio | voice output terminal 31 Audio | voice input terminal 61 Communication I / F part 62 Initial value memory | storage part 63 Limit value memory | storage part 64 Pushbutton switch 65 arithmetic processing unit 66 impulse response estimation unit 67 estimated echo signal generation unit 68 adder 611 network 661 impulse response coefficient calculation unit 662 estimation range limit unit 663 impulse response coefficient storage unit 664 initial setting control unit 672 sampling value storage unit 673 echo signal Estimation calculation unit A Estimated range limit value e Voice input signal e 'Estimated echo signal h Impulse response coefficient r Composite signal s Voice output signal w Echo

Claims (5)

An audio output signal receiver that receives a signal representing sound as an audio output signal;
An audio output terminal for outputting an audio output signal received by the audio output signal receiver;
An audio input terminal for receiving an audio input signal representing sound;
An initial value storage unit that stores in advance an impulse response coefficient representing a correlation between the voice output signal and the voice input signal as an initial impulse response coefficient;
Based on the audio output signal received by the audio output signal receiving unit and the audio input signal received by the audio input terminal, the initial impulse response coefficient stored in the initial value storage unit is set as an initial value. An impulse response coefficient calculation unit for calculating a simple impulse response coefficient;
Estimation that estimates a voice input signal received by the voice input terminal as an estimated echo signal from the voice output signal received by the voice output signal receiving unit and the impulse response coefficient calculated by the impulse response coefficient calculation unit An echo signal generator;
A synthesized signal output unit that outputs a synthesized signal obtained by subtracting the estimated echo signal generated by the estimated echo signal generating unit from the audio input signal received by the audio input terminal ;
A limit value storage unit for preliminarily storing an estimated range limit value indicating a variation range allowed for the initial impulse response coefficient stored in the initial value storage unit of the new impulse response coefficient;
An estimation range limiting unit that adjusts the impulse response coefficient calculated by the impulse response coefficient calculation unit so as to limit the impulse response coefficient within a range of an estimation range limit value stored in the limit value storage unit, Echo cancellation circuit. A receiving unit that receives an initial setting instruction for storing an impulse response coefficient in the initial value storage unit;
When the initial setting instruction is received by the receiving unit, an initial setting control unit that stores the new impulse response coefficient calculated by the impulse response coefficient calculating unit as the initial impulse response coefficient in the initial value storage unit The echo cancellation circuit according to claim 1 , further comprising: The echo cancellation circuit according to claim 1 or 2 ,
An audio output unit that converts an audio output signal output from the audio output terminal into a sound wave;
A sound input unit that converts sound waves into the sound input signal and outputs the sound to the sound input terminal;
A housing that houses the echo cancellation circuit, the audio output unit, and the audio input unit so that a positional relationship between the audio output unit and the audio input unit is fixed;
The impulse response coefficient includes an audio input signal received from the audio input unit by the audio input terminal and the audio output when the audio output signal output from the audio output terminal is converted into a sound wave by the audio output unit. An acoustic device characterized by expressing a correlation with a signal. The echo cancellation circuit according to claim 1 or 2 ,
An audio output unit that converts an audio output signal output from the audio output terminal into a sound wave;
A sound input unit that converts sound waves into the sound input signal and outputs the sound to the sound input terminal;
An image capturing unit for capturing an image by capturing an image;
An image data output unit that transmits image data acquired by the imaging unit to a terminal device connected via a network;
The audio output signal receiving unit receives the audio output signal from the terminal device via the network,
The network camera, wherein the combined signal output unit transmits the combined signal to the terminal device via the network. An audio output signal receiving step of receiving a signal representing sound as an audio output signal;
An audio output step of converting the audio output signal received by the audio output signal reception step into a sound wave;
A voice input process for converting sound waves into a voice input signal;
In the voice output step, when the voice output signal received in the voice output signal reception step is converted into a sound wave, an impulse representing a correlation between the voice input signal obtained in the voice input step and the voice output signal An initial value storage step of storing the response coefficient in advance in the initial value storage unit as an initial impulse response coefficient;
Based on the voice output signal received by the voice output signal receiving step and the voice input signal obtained by the voice input step, with the initial impulse response coefficient stored in the initial value storage unit as the initial value, a new An impulse response coefficient calculation step for calculating a simple impulse response coefficient;
The voice input obtained by the voice input step based on the sound wave output in the voice output step from the voice output signal received in the voice output signal reception step and the impulse response coefficient calculated in the impulse response coefficient calculation step An estimated echo signal estimating step for generating a signal as an estimated echo signal;
A synthesized signal output step of outputting a synthesized signal obtained by subtracting the estimated echo signal generated by the estimated echo signal estimating step from the speech input signal obtained by the speech input step ;
Preliminarily storing in the limit value storage unit an estimated range limit value indicating a variation range allowed for the initial impulse response coefficient stored in the initial value storage unit of the new impulse response coefficient;
An estimation range limiting step of adjusting the impulse response coefficient calculated in the impulse response coefficient calculation step so as to limit the impulse response coefficient within a range of the estimation range limit value stored in the limit value storage unit, How to cancel echo.

Images