JP2013005106A - In-house sound amplification system, in-house sound amplification method, and program therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an in-house sound amplification system which can amplify sound, making it more natural and easier for people to understand than conventional ones, at a near-end side of the inside of a building, and also transmit such sound to a far-end side.SOLUTION: An in-house sound amplification system adds received speech signals from a far-end side and signals based on speech signals acquired by sound acquisition means (amplification speech signals) by an adder to generate amplification signals, and performs voice communication with the far-end side by amplifying the generated amplification signals in a building by amplification means. The in-house sound amplification system comprises: a loss control system including a sound signal detector, a convergence determination device, and a loss controller; and an echo canceler including an echo path estimation device, an artificial echo generator, and a subtracter. The loss control system controls a signal level of the amplification speech signals according to a convergence state of echo path estimation. The echo canceler estimates an acoustic transmission characteristic vector between the amplification means and the sound acquisition means on the basis of the received speech signals from the far-end side and the amplification speech signals, to delete both acoustic echoes derived from the received speech signals from the far-end side and acoustic echoes derived from the amplification speech signals.

Description

  The present invention eliminates the echo that causes the howling and the audibility hindrance in the case of a voice call while simultaneously expanding the reception signal from the far end and the speech signal on the near end side in the near end side field, The present invention relates to an on-site loudspeaker device, an on-site loudspeaker method, and a program thereof.

  A schematic diagram of the on-site loudspeaker is shown in FIG. In a case where a voice call is made between two points, when one is a near end side and the other is a far end side, the in-speech loudspeaker is arranged on the near end side. The in-speech loudspeaker adds a signal obtained by adding the received signal x (n) arriving at the near end side through the network 1 from the far end side and the speech signal s (n) at the near end side by an adder 2 to a speaker or the like. The voice call is performed while the voice is loudened by the loudspeaker 3 (n is a discrete time). At this time, the signal amplified by the loudspeaker 3 also circulates in the sound collecting means 4 such as a microphone, and is collected together with s (n) that should be collected independently. This sneak signal is called acoustic echo, but acoustic echo is a major problem in a telephone system. Specifically, it causes telephone disabilities and discomfort to the caller due to howling. ing.

  Therefore, improvement of the in-speech loudspeaker has been promoted to cope with such a problem. FIG. 13 is a configuration example of the in-field loudspeaker 10 disclosed in Patent Document 1. The in-field loudspeaker 10 includes a loss control system 11, a frequency modulation system 12, and an echo canceller 13 in addition to the adder 2, the loudspeaker 3 and the sound pickup means 4 shown in FIG. The loss control system 11 includes an audio signal detector 11a and a loss controller 11b, the frequency modulation system 12 includes a frequency modulator 12a, and the echo canceller 13 subtracts an echo path estimator 13a and a pseudo echo generator 13b. A container 13c is provided.

The in-field loudspeaker 10 solves the problem of howling caused by the acoustic echo y _s (n) derived from the loud sound from the loudspeaker 3 based on the near-end utterance signal s (n), the loss control system 11 and the frequency modulation. The echo canceller 13 solves the problem of the acoustic echo y _x (n) derived from the loud sound from the loudspeaker 3 of the received signal x (n) from the far end side as well as the system 12.

The loss control system 11 suppresses occurrence of howling when s (n) does not exist. Specifically, when the audio signal detector 11a monitors the transmission signal s ^ (n) to the far end side and determines that s (n) does not exist, s ^ (n) is lost. In this way, the loss controller 11b is operated to control the loudness volume of _s (n) to suppress howling due to y _s (n). On the other hand, the frequency modulation system 12 suppresses occurrence of howling when s (n) exists. Specifically, the frequency modulator 12a applies frequency modulation to s ^ (n) to generate a loud speech signal s ^ '(n) that is loudened from the loudspeaker 3. By adding frequency modulation, it is possible to prevent acoustic echo components from staying in a specific frequency band, so that s ^ '(n) can be loudly loud without causing howling due to y _s (n). .

The echo canceller 13 first uses the echo path estimator 13a to calculate the estimated value H ^ (n) of the acoustic transfer characteristic vector H (n) of the echo path from the loudspeaker 3 to the sound pickup means 4 as x (n). Calculate from s ^ (n). Here, H (n) is a vector (K is an impulse response length) having an impulse response {h ₀ (n), h ₁ (n),..., H _K-1 (n)} as elements. Subsequently, the pseudo echo generator 13b generates a pseudo echo signal y _x ^ (n) simulating y _x (n) from x (n) and H ^ (n). Then, the subtractor 13c subtracts y _x ^ (n) from the sound pickup signal s ′ (n) (= s (n) + y _x (n) + y _s (n)) in the sound pickup means 4 to obtain x. The acoustic echo y _x (n) caused by (n) is eliminated, and a transmission signal s ^ (n) to the far end side is obtained. The detailed operation of the echo canceller 13 is described in Patent Document 1.

Japanese Patent Laid-Open No. 11-340881

In the conventional in-field loudspeaker 10 shown in Patent Document 1, only the acoustic echo y _x (n) derived from the in-field loud sound of the received signal x (n) is eliminated by the echo canceller 13, and the transmitted signal s ^ ( In n), the acoustic echo y _s (n) derived from the in-field loud sound based on the utterance signal s (n) remains superimposed, and it becomes difficult to hear the voice at the far end side receiving s ^ (n). There is a problem. Further, the in-field loudspeaker 10 generates the loud utterance signal s ^ '(n) to be amplified in the field by performing frequency modulation on the utterance signal s (n). There is also a problem that the sound becomes natural.

  It is an object of the present invention to provide an in-field loudspeaker, an in-field loudspeaker method, and a program thereof that can louden natural and easy-to-hear sound in the field on the near end side and that can also be transmitted to the far end side. It is to provide.

  The in-field loudspeaker of the present invention includes an audio signal detector, a convergence determiner, a loss controller, an echo path estimator, a pseudo echo generator, and a subtracter. The audio signal detector determines a transmission / reception state of a transmission signal to the far end side and a reception signal received from the far end side. The convergence determiner includes a storage unit, and when the speech signal detector determines that the received signal is present, determines a convergence state of the echo path estimation by a predetermined method, and the convergence of the echo path estimation proceeds. The amount of loss is calculated so as to be as small as possible, and this is stored in the storage unit and provided to the loss controller, and when the voice signal detector does not determine that the received signal is present, The loss amount stored in the storage unit is provided to the loss controller. The loss controller uses the transmission signal and generates a loud utterance signal by giving a loss corresponding to the loss amount provided from the convergence determination unit. The echo path estimator outputs a sound transmission characteristic vector of an echo path from the sound enhancement means to the sound collection means, and a signal (speeched from the sound enhancement means, which is the sum of the transmission signal, the received signal, and the loud speech signal ( Hereinafter, the estimated sound transfer characteristic vector (hereinafter referred to as “pseudo sound transfer characteristic vector”) is output. The pseudo echo generator generates a pseudo echo signal by using the loud sound signal and the pseudo acoustic transfer characteristic vector. The subtracter generates the transmission signal by subtracting the pseudo echo signal from a sound collection signal including the speech signal collected by the sound collection means.

  An in-field loudspeaker, an in-field loudspeaker method, and a program thereof according to the present invention are obtained by frequency-modulating a loudspeaked speech signal based on a speech signal included in a loudspeaker signal loudened by a loudspeaker to a transmission signal as in the prior art. If the received signal is not generated, it is generated by giving a smaller loss to the outgoing signal to the far-end side as the convergence of the echo path estimation progresses, and if the received signal does not exist Generate by giving the loss generated in the presence. In this way, by performing loss control in accordance with the convergence state, howling is suppressed, and speech that is more natural and easier to hear than before can be expanded into the near-end side field. Further, when obtaining the pseudo-acoustic transfer characteristic vector for echo path estimation, conventionally, only the transmission signal and the reception signal to the far end side are used. However, in the present invention, a loud speech signal is also used. Thereby, when calculating the pseudo acoustic transfer characteristic vector, not only the acoustic echo derived from the loud sound from the loudspeaker of the received signal but also the acoustic echo derived from the loud sound from the loudspeaker based on the speech signal is considered. Therefore, pseudo echo signals for both acoustic echoes can be generated. Therefore, both acoustic echoes can be eliminated from the collected sound signal, and a natural and easy-to-hear signal can be sent to the far end side.

The figure which shows the structural example of the in-speech loudspeaker 100. FIG. The figure which shows the example of a processing flow of the in-field loudspeaker 100. The figure which shows the calculation image of the loss amount based on the determination result of the convergence state in a convergence determination device. The figure which shows the structural example of the in-speech loudspeaker 200. FIG. The figure which shows the structural example of the inside loudspeaker 300. FIG. The figure which shows the structural example of the loudspeaker 400 in a hall. The figure which shows the structural example of the inside loudspeaker 500. FIG. The figure which shows the structural example of the loudspeaker 600 in a hall. The figure which shows the structural example of the inside loudspeaker 700. FIG. The figure which shows the structural example of the on-site loudspeaker 800. The figure which shows the structural example of the inside loudspeaker 900. The figure which shows a loudspeaker in a field typically. The figure which shows the structural example of the conventional loudspeaker 10 in a field.

  Hereinafter, embodiments of the present invention will be described in detail.

  FIG. 1 shows an example of the configuration of the on-site loudspeaker 100 of the present invention, and FIG. 2 shows an example of its processing flow. When performing a voice call between two points, the in-speech loudspeaker 100 is arranged on the near end side when one is the near end side and the other is the far end side, and the received signal x ( n) and the loud speech signal s ^ '(n) based on the speech signal s (n) picked up by the sound pickup means 4 on the near end side are added by the adder 2 to generate the loud sound signal x (n ) + S ^ '(n) is an in-field loudspeaker that makes a voice call with the far-end side while loudening the near-end side with the loudspeaker 3. Here, n is a discrete time, and is set, for example, at an interval having a value smaller than an interval (about 1.66 msec) at which a human voice band can be expressed by a discrete time signal.

  The in-field loudspeaker 100 includes a loss control system 110 and an echo canceller 130 in addition to the adder 2, the loudspeaker 3 and the sound pickup means 4.

The loss control system 110 includes a speech signal detector 111, a convergence determination unit 112, and a loss controller 113, and a loud sound (a loud speech signal s ^ '(n) from the loudspeaker 3 based on the speech signal s (n). ) To solve the howling problem caused by the acoustic echo y _s (n).

The audio signal detector 111 determines the transmission / reception state of the reception signal x (n) and the transmission signal s ^ (n) to the far end side (S1). The transmission / reception state can be determined by the method described in Reference 1, for example.
[Reference 1] Japanese Patent Laid-Open No. 11-331046

The convergence determination unit 112 includes a storage unit 112a (not shown). When the speech signal detector 111 determines that the received signal x (n) is present, the convergence determination unit 112 determines the convergence state of the echo path estimation by a predetermined method. To do. For example, the convergence state of the echo path estimation is determined from the signal s ^ (n) transmitted to the far end side and the sound collection signal s ′ (n) collected by the sound collection means 4. In the collected sound signal s ′ (n), in addition to the utterance signal s (n), an acoustic echo y _x (n) derived from the loud sound (x (n)) from the loud sound means 3 of the received signal _x (n). And the sound echo y _s (n) derived from the loud sound from the loud sound means 3 based on the speech signal s (n) (the loud speech signal s ^ ′ (n)). Further, the transmission signal s ^ (n) to the far end side is obtained by subtracting the pseudo echo signal y ^ (n) generated by the pseudo echo generator 132 from the collected sound signal s' (n). Based on the determination result of the convergence state, the loss amount is calculated so as to become a smaller value as the convergence of the echo path estimation progresses, and is stored in the storage unit 112a and provided to the loss controller 113. On the other hand, if the voice signal detector 111 does not determine that the received signal x (n) is present, the loss amount stored in the storage unit 112a is provided to the loss controller 113 (S2).

  The loss controller 113 uses the transmission signal s ^ (n) and gives a loss of the loss amount provided from the convergence determination unit 112 to generate a loud speech signal s ^ '(n) (S3). . A loud voice signal x (n) + s ^ '(n) generated by adding the loud voice signal s ^' (n) and the received signal x (n) generated here by the adder 2 is loud voice. Aloud from the means 3 into the hall.

  The calculation of the loss amount based on the determination result of the convergence state in the convergence determiner 112 and the loss control in the loss controller 113 are performed as follows, for example. Fig. 3 (a) shows an image of calculating the amount of loss.

  When the transmission signal s ^ (n) is hardly smaller than the sound pickup signal s ′ (n), for example, s ^ (n) / s ′ (n) is a predetermined first threshold (for example, 0.9) or more. If it is the value of, it is determined that the convergence has hardly progressed, and a magnification value of, for example, about 0 to 0.3 is calculated as the loss amount so that the loss given to s ^ (n) becomes large. The loss controller 113 multiplies s ^ (n) by this magnification value to generate a loud speech signal s ^ '(n).

  When the transmission signal s ^ (n) is sufficiently smaller than the sound collection signal s '(n), for example, the level ratio s ^ between the transmission signal s ^ (n) and the sound collection signal s' (n) When (n) / s ′ (n) is a value equal to or smaller than a predetermined second threshold (for example, 0.7), it is determined that the convergence is sufficient and the loss given to s ^ (n) is reduced. For example, a magnification value of about 0.7 to 1.0 is calculated as the loss amount. The loss controller 113 multiplies s ^ (n) by this magnification value to generate a loud speech signal s ^ '(n).

  In other cases, that is, when s ^ (n) / s '(n) is a value between the first threshold value and the second threshold value, for example, s ^ (n) / s' (n) For example, a magnification value of about 0.3 to 0.7 is calculated as the loss amount so that the loss given to s ^ (n) monotonously decreases as the value decreases. The loss controller 113 multiplies s ^ (n) by this magnification value to generate a loud speech signal s ^ '(n).

  In addition, although the method of determining whether the convergence is sufficient or not in three stages using a predetermined threshold value for s ^ (n) / s' (n) is illustrated here, the determination stage may be two stages or four stages. That's fine. Further, the calculation of the loss amount is not necessarily stepwise, and for example, it may be calculated as a continuous value given by a function that monotonously decreases as the convergence progresses.

  As described above, in the present invention, the loudspeaked speech signal s ^ '(n) based on the speech signal included in the loudspeaker signal loudened by the loudspeaker is frequency-modulated to the transmission signal s ^ (n) as in the conventional case. When the received signal x (n) exists, the generated signal is generated by giving a smaller loss to s ^ (n) as the convergence of the echo path estimation progresses, and the received signal x If (n) does not exist, it is generated by giving the loss generated when it exists. In this way, by performing loss control in accordance with the convergence state, howling is suppressed, and speech that is more natural and easier to hear than before can be expanded into the near-end side field.

The echo canceller 130 includes an echo path estimator 131, a pseudo echo generator 132, and a subtracter 133, and a loud sound from the loud sound means 3 (a loud speech signal s ^ '(n)) based on the speech signal s (n). The problem of the acoustic echo y _x (n) derived from the loud sound (x (n)) from the loudspeaker 3 of the received signal x (n) and the acoustic echo y _s (n) derived from the sound is solved.

The echo path estimator 131 obtains the estimated value H ^ (n) of the acoustic transfer characteristic vector of the echo path from the loudspeaker means 3 to the sound pickup means 4, and outputs the transmitted signal s ^ (n) and the loudspeaker signal x (n) + s ^ It calculates | requires using '(n) (S4). Here, the estimated value H (n) of the acoustic transfer characteristic vector is a vector whose elements are impulse responses {h ₀ (n), h ₁ (n),... H _K-1 (n)} (K is an impulse. Response length), hereinafter referred to as a pseudo acoustic transfer characteristic vector.

  The pseudo echo generator 132 generates the pseudo echo signal y ^ (n) using the loud sound signal x (n) + s ^ '(n) and the pseudo acoustic transfer characteristic vector H ^ (n) (S5).

  The subtracter 133 generates the transmission signal s ^ (n) by subtracting the pseudo echo signal y ^ (n) from the collected sound signal s' (n) (S6).

  Since other detailed operations of the echo canceller 130 are described in Patent Document 1, description thereof will be omitted.

In obtaining H ^ (n) for echo path estimation, conventionally, only the transmitted signal s ^ (n) and the received signal x (n) are used. However, in the present invention, the voiced speech signal s ^ 'is further increased. (n) is also used. Thus, in estimating H ^ (n), not only the acoustic echo y _x (n) derived from the loud sound (x (n)) from the loud sound means 3 of the received signal _x (n) but also the speech signal s ( Since the acoustic echo y _s (n) derived from the loud sound from the loudspeaker 3 based on (n) is also considered, the acoustic echo y (n) = y _x (n ) + Y _s (n), a pseudo echo signal y ^ (n) can be generated. For this reason, both acoustic echoes y _x (n) and y _s (n) can be eliminated from the collected sound signal s ′ (n), and the transmission signal s ^ (n), which is more natural and easier to hear than before, can be obtained on the far end side. Can be sent to.

  FIG. 4 shows a configuration example of the on-site loudspeaker 200 of the present invention. The on-site loudspeaker 200 includes an adder 2, loudspeaker 3, sound pickup means 4, loss control system 210, and echo canceller 130. The loss control system 210 includes an audio signal detector 111, a convergence determination unit 212, and a loss controller 113, and the echo canceller 130 includes an echo path estimator 131, a pseudo echo generator 132, and a subtracter 133. In other words, only the convergence determination unit 212 is different from the in-field loudspeaker 100 of the first embodiment, and other devices with the same reference numerals are the same. In the following, the same reference numerals are assigned to the same devices, explanations are omitted in principle, and touched as necessary.

  The convergence determination unit 212 determines the convergence state of the echo path estimation from the signal level of the transmission signal s ^ (n) to the far end side and the signal level stored in the storage unit 112a. In addition, the signal level preserve | saved at the memory | storage part 112a shall be preserve | saved based on a predetermined rule. For example, it is assumed that a negative value (for example, −∞) as small as possible is stored as the initial value of the signal level, and the stored signal level and the transmission signal s ^ (n) are sequentially compared to determine the signal level. A rule of sequentially updating the stored signal level on the larger side is conceivable.

  The calculation of the loss amount based on the determination result of the convergence state in the convergence determination unit 212 is performed as follows, for example. Fig. 3 (b) shows an image of calculating the amount of loss.

  When the signal level of the send signal s ^ (n) is compared with the stored signal level and the signal level of the send signal s ^ (n) is large, for example, “s ^ (n) / stored signal level” is When the value is equal to or greater than a predetermined first threshold (for example, 0.9), it is determined that the convergence has not substantially progressed, and the loss amount is, for example, about 0 to 0.3 so that the loss given to s ^ (n) increases. The magnification value is calculated.

  When the signal level of the transmission signal s ^ (n) is sufficiently smaller than the stored signal level, for example, “s ^ (n) / stored signal level” is a predetermined second threshold (for example, 0.7). If it is the following value, it is determined that it has converged sufficiently, and a magnification value of, for example, about 0.7 to 1.0 is calculated as the loss amount so that the loss given to s ^ (n) becomes small.

  In other cases, that is, when “s ^ (n) / stored signal level” is a value between the first threshold value and the second threshold value, for example, “s ^ (n) / stored signal level”. For example, a magnification value of about 0.3 to 0.7 is calculated as the loss amount so that the loss given to s ^ (n) monotonously decreases as the value decreases.

  FIG. 5 shows a configuration example of the on-site loudspeaker 300 of the present invention. The on-site loudspeaker 300 includes an adder 2, loudspeaker 3, sound pickup means 4, loss control system 310, and echo canceller 130. The loss control system 310 includes an audio signal detector 111, a convergence determination unit 312, and a loss controller 113, and the echo canceller 130 includes an echo path estimator 131, a pseudo echo generator 132, and a subtracter 133. That is, only the convergence determination unit 312 is different from the on-site loudspeaker 100 of the first embodiment.

  The convergence determination unit 312 determines the convergence state of the echo path estimation from the magnitude of the pseudo acoustic transfer characteristic vector H ^ (n) and the magnitude of the past pseudo acoustic transfer characteristic vector stored in the storage unit 112a. It is assumed that the past pseudo-acoustic transfer characteristic vectors stored in the storage unit 112a are stored based on a predetermined rule. For example, assume that H ^ (n−x) is stored. x may be an arbitrary value of 1 or more, but if the echo path is changed due to the movement of the apparatus or the movement of the person, the accuracy of convergence determination deteriorates. Therefore, it is desirable that the value be close to the current time, that is, as small as possible.

  The calculation of the loss amount based on the determination result of the convergence state in the convergence determination unit 312 is performed as follows, for example. Fig. 3 (c) shows an image of calculating the loss amount.

  If the two vectors do not agree well, for example, if | H ^ (n) / H ^ (n−x) | is a value smaller than a predetermined first threshold value (for example, 0.7), the convergence is almost advanced. For example, a magnification value of about 0 to 0.3 is calculated as the loss amount so that the loss given to s ^ (n) becomes large.

  If both vectors match well, for example, if | H ^ (n) / H ^ (n−x) | is a value of a predetermined second threshold (for example, 0.95 to 1.05), sufficient convergence is achieved. As a loss amount, for example, a magnification value of about 0.7 to 1.0 is calculated so that the loss given to s ^ (n) is small.

  In other cases, that is, when | H ^ (n) / H ^ (nx) | is a value between the first threshold value and the second threshold value, for example, | H ^ (n) / H For example, a magnification value of about 0.3 to 0.7 is calculated as the loss amount so that the loss given to s ^ (n) monotonously decreases as the value of ^ (n−x) | increases.

  A configuration example of the on-site loudspeaker 400 of the present invention is shown in FIG. The on-site loudspeaker 400 includes an adder 2, loudspeaker 3, sound collection unit 4, loss control system 410, and echo canceller 130. The loss control system 410 includes an audio signal detector 111, a convergence determination unit 412, and a loss controller 113, and the echo canceller 130 includes an echo path estimator 131, a pseudo echo generator 132, and a subtracter 133. In other words, only the convergence determination unit 412 is different from the in-field loudspeaker 100 of the first embodiment.

The convergence determination unit 412 uses the determination methods of the convergence determination units 112, 212, and 312 in combination. That is, the ratio between the signal level of the transmission signal s ^ (n) representing the convergence state of the echo path estimation and the signal level of the sound pickup signal s' (n), and the signal level s ^ (n) of the transmission signal The ratio of the signal level stored in the storage unit based on a predetermined rule, the size of the pseudo acoustic transfer characteristic vector H ^ (n), and the past pseudo sound stored in the storage unit based on the predetermined rule For two or more ratios with respect to the size of the transfer characteristic vector, a convergence state is determined based on the ratio value, and the loss amount is calculated. That is, when determining by all three ratios, three loss amounts C ₁ , C ₂ , and C ₃ corresponding to them are respectively calculated, and when determining by any two ratios, they correspond to them. Two loss amounts C ₁ and C ₂ are respectively calculated. Then, for example, a value obtained by weighted averaging these loss amounts is set as a loss amount to be given to the loss controller 113, and is stored in the storage unit 112a.

  The in-field loudspeaker 500 according to the fifth embodiment includes a loss controller 113 and an adder 2 in the configurations of the first to fourth embodiments in order to suppress the occurrence of howling due to in-field loudspeaking without substantially changing the audibility. The frequency domain loss controller 514 is inserted between them. FIG. 7 shows an example in which the in-field loudspeaker 500 is configured by adding a frequency domain loss controller 514 to the in-field loudspeaker 400 of the fourth embodiment.

  First, the frequency domain loss controller 514 uses the pseudo-acoustic transfer characteristic vector H ^ (n) generated by the echo path estimator 131 and / or the loud speech signal s ^ '(n) generated by the loss controller 113. Frequency conversion is performed, and a frequency band whose power is equal to or higher than a predetermined value (for example, 0.7 * 0.7 = 0.49 times the square of the maximum output value 0.7) is identified as a frequency band in which howling may occur. Then, a predetermined loss (for example, 0.5 times) is applied to the voiced speech signal s ′ ′ (n) generated by the loss controller 113 using a notch filter (band elimination filter) or the like only in the specified frequency band. Loss) is given, and the speech signal s ^ '(n) given this loss is given to the adder 2.

  It should be noted that the frequency band that causes loss may be one frequency band that is most likely to generate howling among the frequency bands that are determined to have the possibility of howling, and the possibility of howling occurring. The frequency band may be all of a certain frequency band, or may be limited to a preset number of frequency bands.

  The on-site loudspeaker 600 according to the sixth embodiment has a configuration in which the function of the frequency domain loss controller 514 described in the fifth embodiment is integrated with the loss controller, and can be applied to any of the first to fourth embodiments. FIG. 8 shows an example in which the in-field loudspeaker 600 is configured by replacing the loss controller 113 of the in-field loudspeaker 400 of the fourth embodiment with a loss controller 613 having the function of the frequency domain loss controller 514.

  The loss controller 613 uses the transmission signal s ^ (n), first converts it to the frequency domain, and then gives the loss amount provided by the convergence determination unit to all frequency bands. Subsequently, for the pseudo acoustic transfer characteristic vector H ^ (n) converted into the frequency domain and / or the frequency domain transmission signal s ^ (n) given the loss, the frequency band in which the power is a predetermined value or more is specified. . Then, the transmission signal s ^ (n) in the frequency domain where the loss is given is subjected to a predetermined loss (for example, about 0.5 times larger) by using a notch filter (band elimination filter) or the like only in the specified frequency band. Loss), which is converted to the time domain, is output as a loud speech signal s ^ '(n).

  The in-speech loudspeaker 700 according to the seventh embodiment has a configuration in which a delay inserter 715 is inserted before the adder 2 in the configurations of the first to sixth embodiments to give a predetermined delay to the loudspeaking speech signal s ^ '(n). is there. FIG. 7 shows an example in which the in-field loudspeaker 700 is configured by adding a delay inserter 715 to the in-field loudspeaker 400 of the fourth embodiment.

  It is known that a speaker adjusts an utterance while listening to a voice uttered by the speaker. Among them, with regard to the delay time, it has been confirmed by experiments that it feels easy to speak when listening to a voice uttered by itself with a delay of about 0 msec to 40 msec. In particular, it has been found that it is easiest to speak when one's own speech is heard with a delay of about 20 msec to 30 msec.

  Therefore, the delay time is determined so that the sum of the delay generated when the voice travels through the space and the delay generated in the system is 0 msec to 40 msec (preferably 20 msec to 30 msec). The delayed inserter 715 gives the loud speech signal s ^ '(n). This makes it easier for the near-end speaker to speak.

  The in-field loudspeaker 800 according to the eighth embodiment, when the convergence of the echo path estimation is sufficient in the convergence determination device and the loss controller according to the first to seventh embodiments, The loudspeaker 3 can be controlled so as to have a predetermined volume when the loudspeaker 3 performs loudspeaking.

  As described in the seventh embodiment, it is known that the speaker adjusts the utterance while listening to the volume of the uttered voice. Therefore, if the convergence of the echo path estimation is already sufficient, the volume that makes it easier for the near-end speaker to speak (for example, the average sound pressure at the speaker's ear when the speaker continues speaking) The loss controller is controlled so that the in-field loud sound can be presented to the speaker at a level of 55 dB.

  FIG. 10 shows an example in which the in-field loudspeaker 800 is configured by replacing the convergence determiner 412 and the loss controller 113 of the in-field loudspeaker 400 of the fourth embodiment with the convergence determiner 812 and the loss controller 813 of the present embodiment. .

  When the convergence of the echo path estimation is sufficient, the convergence determination unit 812 is a loss controller so that the volume of the loud speech signal s ^ '(n) becomes a predetermined volume when the loudspeaker 3 performs loudspeaking. 813 is controlled. The loss controller 813 adjusts the volume of the loud speech signal s ^ ′ (n) according to the control from the convergence determination unit 812.

  The on-site loudspeaker 900 according to the ninth embodiment is configured to perform processing in each device in the frequency domain in the first to eighth embodiments. By performing processing in the frequency domain, the amount of calculation can be reduced, and in combination with processing such as making the discrete time finer, in-field loudspeaking with more natural sound becomes possible.

  FIG. 11 shows an example in which the on-site loudspeaker 900 is configured by adding the frequency domain transforming units 941 and 943 and the time domain transforming units 942 and 944 of this embodiment to the on-site loudspeaker 400 of the fourth embodiment.

  The frequency domain transformer 941 converts the received signal x (n) in the time domain into the received signal x (ω) in the frequency domain, and the frequency domain transformer 943 collects the collected sound signal s ′ (n) in the time domain in the frequency domain. Each is converted into a sound signal s ′ (ω).

  The loss control system 410 and the echo canceller 130 perform processing in the frequency domain.

  The time domain transforming unit 942 converts the frequency domain loudspeaker signal x (ω) + s ^ ′ (ω) into the time domain loudspeaker signal x (n) + s ^ ′ (n), and the time domain transforming unit 944 operates the frequency domain. The transmission signal s ^ (ω) is converted into a time-domain transmission signal s ^ (n).

  Each process in the in-field loudspeaker and the in-field loudspeaker of each embodiment described above is performed not only in time series according to the description, but also in parallel or individually according to the processing capability of the apparatus that performs the process or as necessary. May be executed. Further, the functions of the in-situ loudspeaker of the present invention may be merged and divided as necessary. Other modifications can be made as appropriate within the scope of the technical idea expressed in the present invention.

  When the equipment constituting the on-site loudspeaker of the present invention is realized by a computer, the processing contents of the functions of each equipment are described by a program. The program is stored in, for example, a hard disk device, and at the time of execution, necessary programs and data are read into a RAM (Random Access Memory), and the program is executed by the CPU. The content is realized.

Claims (13)

A voice signal detector for determining a transmission / reception state of a transmission signal to the far end side and a reception signal received from the far end side;
A storage unit, and when the speech signal detector determines that the received signal is present, determines a convergence state of the echo path estimation by a predetermined method, and the smaller the value as the convergence of the echo path estimation progresses, The amount of loss is calculated and stored in the storage unit and provided to the loss controller. If the voice signal detector does not determine that the received signal is present, the loss amount is stored in the storage unit. A convergence determiner that provides the loss controller
A loss controller that uses the transmission signal and generates a loud speech signal by giving a loss of a loss amount provided from a convergence determiner;
The sound transmission characteristic vector of the echo path from the sounding means to the sound collecting means is a signal (hereinafter referred to as “sounding signal”) that is a sound amplified from the sounding means that is the sum of the transmission signal, the received signal, and the sounding speech signal. And an echo path estimator that outputs an estimated acoustic transfer characteristic vector (hereinafter referred to as “pseudo-acoustic transfer characteristic vector”),
A pseudo echo generator that generates a pseudo echo signal using the loud sound signal and the pseudo acoustic transfer characteristic vector;
A subtractor for generating the transmission signal by subtracting the pseudo echo signal from a sound pickup signal including a speech signal picked up by the sound pickup means;
An on-site loudspeaker. The in-speech loudspeaker according to claim 1,
The convergence determining unit determines a convergence state of echo path estimation in n stages (n is an integer of 2 or more) using a predetermined threshold, and calculates a loss amount corresponding to the stage. apparatus. The in-speech loudspeaker according to claim 1 or 2,
The in-field loudspeaker characterized in that the convergence determination unit obtains a value indicating a convergence state of echo path estimation as a ratio between a signal level of the transmission signal and a signal level of the sound pickup signal. The in-speech loudspeaker according to claim 1 or 2,
The convergence determination unit obtains a value indicating a convergence state of echo path estimation as a ratio between a signal level of the transmission signal and a signal level stored in the storage unit based on a predetermined rule. Loudspeaker. The in-speech loudspeaker according to claim 1 or 2,
The convergence determination unit calculates a value indicating a convergence state of the echo path estimation between the magnitude of the pseudo acoustic transfer characteristic vector and the magnitude of the past pseudo acoustic transfer characteristic vector stored in the storage unit based on a predetermined rule. An on-site loudspeaker characterized by obtaining a ratio. The in-speech loudspeaker according to claim 1 or 2,
The convergence determination unit represents the value indicating the convergence state of the echo path estimation, the ratio between the signal level of the transmission signal and the signal level of the collected sound signal, the signal level of the transmission signal and a predetermined rule. The ratio of the signal level stored in the unit and the ratio of the size of the pseudo acoustic transfer characteristic vector and the size of the past pseudo acoustic transfer characteristic vector stored in the storage unit based on a predetermined rule, For two or more, the convergence state is determined based on the ratio value to calculate the loss amount, and the weighted average value of each calculated loss amount is stored in the storage unit and provided to the loss controller An on-site loudspeaker characterized by: The in-speech loudspeaker according to any one of claims 1 to 6,
Amplifying speech generated by the loss controller by frequency-converting the pseudo-sound transfer characteristic vector and / or the speech signal generated by the loss controller and specifying a frequency band whose power is a predetermined value or more. An in-field loudspeaker further comprising a frequency domain loss controller that gives a predetermined loss only to the specified frequency band to a signal and outputs the loss as a loud speech signal constituting the loud signal. The in-speech loudspeaker according to any one of claims 1 to 6,
The loss controller uses the transmission signal, converts it into the frequency domain, gives the loss of the loss amount provided from the convergence determination unit to all frequency bands, and converts the pseudo acoustic transmission into the frequency domain. A frequency band whose power is equal to or higher than a predetermined value is specified for the transmission signal in the frequency domain given the loss by specifying a frequency band whose power is a predetermined value or more for the transmission signal in the frequency domain given the characteristic vector and / or the loss. An in-field loudspeaker characterized in that a predetermined loss is given only to a specific frequency band, and this is converted into the time domain and output as the loud speech signal. The in-speech loudspeaker according to any one of claims 1 to 8,
An in-field loudspeaker further comprising a delay inserter that gives a predetermined delay to the loud voice signal and outputs the delayed voice signal as a loud voice signal constituting the loud voice signal. The in-speech loudspeaker according to any one of claims 1 to 9,
When the convergence of the echo path estimation is sufficient, the convergence determiner controls the loss controller so that the volume of the loud utterance signal becomes a predetermined volume when the loudspeaker is loudened,
The in-field loudspeaker characterized in that the loss controller adjusts the volume of the loud-speaking speech signal in accordance with control from the convergence determination unit. The in-speech loudspeaker according to any one of claims 1 to 10,
A frequency domain converter that converts the received signal and the collected sound signal into a frequency domain, respectively;
A time domain converter for converting the loudspeaker signal and the transmission signal in the frequency domain to the time domain respectively;
Further comprising
An on-site loudspeaker characterized in that processing in each device is performed in the frequency domain. A voice signal detection step for determining a transmission / reception state of a transmission signal to the far end side and a reception signal received from the far end side;
If it is determined in the audio signal detection step that the received signal is present, the convergence state of the echo path estimation is determined by a predetermined method, and the amount of loss is reduced as the convergence of the echo path estimation progresses. Is calculated and stored in the storage unit and provided to the loss control step. If it is not determined that the received signal is present in the audio signal detection step, the loss amount stored in the storage unit is lost. A convergence determination step provided to the control step;
A loss control step of generating a loud speech signal by using the transmission signal and giving a loss of the loss amount provided from the convergence determination step to the transmission signal;
The sound transmission characteristic vector of the echo path from the sounding means to the sound collecting means is a signal (hereinafter referred to as “sounding signal”) that is a sound amplified from the sounding means that is the sum of the transmission signal, the received signal, and the sounding speech signal. And an echo path estimation step for outputting an estimated acoustic transfer characteristic vector (hereinafter referred to as a “pseudo-acoustic transfer characteristic vector”),
A pseudo echo generating step for generating a pseudo echo signal using the loud sound signal and the pseudo acoustic transfer characteristic vector;
A subtracting step for generating the transmission signal by subtracting the pseudo echo signal from a sound collection signal including an utterance signal collected by the sound collection means;
Perform in-speak loudspeaker method.   A program for causing a computer to function as the on-site loudspeaker according to any one of claims 1 to 11.

Images