JP2002034092A - Sound-absorbing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a sound absorbing device which collects only a target sound at a high sound-collecting S/N ratio, even in a noise by a method where a sound source is specified, without performing a sound-collecting scanning operation and to control its directivity. SOLUTION: The sound-collecting device is provided with a sound-collecting part 10 with directivity, a directivity control part 30, a selection part 40 which selects a target sound source region by a detection key word and a key-word detection part 60 by a voice recognition means. The key word and the sound source region are registered in advance, so as to be related to a reference table 50. When the key word is detected by the part 60, the related sound source region is selected, and the directivity of the part 10 is controlled toward the selected target sound source region in synchronization with a signal from a control-timing instruction part 20.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a sound pickup device having directivity and capable of controlling the directivity. In particular, the present invention relates to a device that eliminates noise and collects only a target sound at a high SN ratio by specifying a target sound source region and controlling the directivity according to the sound source region.

[0002]

2. Description of the Related Art In an environment where a lot of noises exist, such as in a room where people gather or in a car, in order to collect sound with a high SN ratio, a close-talking microphone is conventionally required. However, preparing the close-talking microphones for the number of speakers to collect sound is not only troublesome for the speaker to use, but also increases the cost and is inferior in system expandability. Therefore, if the sound pickup unit has directivity and the directivity can be appropriately directed to the sound source, a high sound pickup S / N ratio can be obtained even if the sound pickup unit is arranged at a position distant from the speaker. A natural conversation that is not aware of the existence of is possible.

A typical example of a sound collection device having a sound collection unit having directivity and capable of controlling the directivity is a microphone array device.

The microphone array device performs signal separation and noise removal by arranging a plurality of microphones at spatially different positions and performing signal processing. Microphone array devices require complex signal processing, but the development of high-performance algorithms and the improvement of processor processing power have realized miniaturization and cost reduction. Currently, hands-free communication devices, multimedia communication conference systems Applications are expanding in a wide range of fields.

For example, a system for estimating a target sound source direction distance using a microphone array device for a multimedia communication conference, and performing sound pickup, noise removal, and camera focus control toward the sound source direction has been commercialized. Have been.

As such an example, Japanese Patent Application Laid-Open No. 9-2512
Japanese Patent Publication No. 99 is a microphone array input type speech recognition device having a speech input unit, a frequency analysis unit, a sound source location search unit, a speech parameter extraction unit, and a speech recognition unit, and is input by the speech input unit. The input signal from the microphone array is frequency-analyzed by a frequency analysis unit using a band-pass filter, and a band-pass waveform for each frequency band is obtained for each microphone channel. From this band-pass waveform, the sound source position or direction is detected by a sound source position search unit. The band-pass power distribution for each frequency band is obtained for each frequency band, and the sound source position or direction is estimated from the band-pass power distribution. The corresponding bandpass power is extracted as a speech parameter from It has been shown that used to identify.

Japanese Unexamined Patent Application Publication No. 11-18192 discloses a sound collection device having a microphone array, a microphone array main device, a speaker, a reception detector, and a directivity controller. It shows that the microphone array main unit performs signal processing on the obtained signal, and directs the microphone array to the area where the target speaker is present to collect the target voice with a high SN ratio.

These devices are characterized in that the microphone array device scans the surroundings and collects sound, and as a result, detects the angle and position where the output from the microphone array main device is the largest, and specifies the angle and position as the position of the speaker. .

[0009]

When a microphone array device picks up sound to specify a sound source, various noises often become obstacles. If the noise is other than human speech, some removal may be possible based on the speech spectrum. However, in a communication conference system, for example, human voice such as a voice from a receiving speaker or a covert voice of a speaker who does not want to collect sound is present as noise. Therefore, the sound signal is collected with high accuracy,
Separation is difficult and has limitations.

Normally, in an official place such as a conference, the position of a speaker is predetermined and basically does not move from there. Another feature is that remarks are controlled depending on the context of the conversation, such as nomination by the chairperson.

The present invention focuses on the fact that the physical position of the speaker is fixed and the context dependency when the speaker changes as described above, and picks up a sound source that is easily affected by noise. An object is to specify a sound source region to be picked up without scanning, control the directivity of a microphone, and pick up a target sound at a high SN ratio.

[0012]

According to the sound pickup apparatus of the present invention,
As means for specifying a target sound source area without performing sound pickup scanning, a keyword input from the outside is detected. That is, the sound pickup device of the present invention includes a sound pickup unit having directivity, a unit for controlling the directivity of the sound pickup unit, a unit for selecting a target sound source area for sound pickup, and a keyword input unit. And
When a keyword is associated with a sound source area candidate obtained in advance, when the keyword is input from the keyword input unit, the associated sound source area is selected, and the sound pickup unit is directed in accordance with the selected area. It is characterized by controlling the characteristics.

According to the present invention, in the above-described sound collecting device, a speech signal which is always picked up as a sound collecting device is processed and used as a keyword input. That is, the present invention provides the above sound collecting apparatus, comprising: means for recognizing a sound signal obtained from a sound collecting unit to obtain sound information; and means for detecting a keyword from the sound information.

Further, according to the present invention, in the sound pickup device, the sound source area candidates associated with the keywords are stored in the device in the form of a comparison table. Since this keyword should be changed according to the situation, the comparison table is stored in a rewritable storage area.

[0015] In the above sound pickup device of the present invention, the keyword is an identifier such as a name of a speaker for identifying a sound source.

The sound collecting apparatus of the present invention includes means for detecting a state of no sound from a sound signal obtained from the sound collecting unit. If the state of no sound continues for a predetermined time or more, the speech is terminated. Judgment is made, and in synchronism therewith, directivity control of the sound pickup unit is performed in accordance with the target sound source area, and the sound of the next speaker is picked up at a high SN ratio.

According to the sound collecting apparatus of the present invention, the directivity of a sound collecting section having directivity is controlled toward a target sound source area.
Keyword detection means for detecting a keyword specifying the next target sound source from the voice input from the sound pickup unit, a keyword classification list storing data indicating the change of the target sound source, and data in which the target sound source and the sound source region are previously associated And a sound source area selecting unit for selecting a sound source area by referring to the keyword-sound source area comparing table using data on the target sound source detected by the keyword detecting means, and a directivity of the sound collecting unit. Control timing instructing means for outputting a control start instruction signal for instructing the timing to start the control to shift to the selected sound source region,
A directivity control unit is provided for controlling the directivity of the sound pickup unit according to the sound source area selection signal and the control start instruction signal.

Further, in the sound pickup apparatus according to the present invention, the control timing instructing means includes means for detecting a state of no sound from a sound signal obtained from the sound collecting section, and the soundless state continues for a predetermined time or more. The control start instruction signal is output in synchronization with the case.

[0019]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing a configuration example of a signal processing unit of a target sound source area designating sound pickup apparatus by keyword detection according to the present invention. The signal processing unit of the target sound source area designating sound collecting device by keyword detection according to the present invention includes a sound collecting unit 10 including a microphone array 11 and a microphone array main device 12, a control timing instructing unit 20, and a directivity control unit 3.
0, a sound source area selection unit 40, a keyword-sound source area comparison table 50, a speech recognition / keyword detection unit 60, and a keyword classification list table 70.

A sound collection unit 10 comprising a microphone array 11 and a microphone array main unit 12 controls a plurality of microphones provided in the microphone array, and emphasizes and extracts only sound in a target sound source area.

The control timing instructing section 20 monitors a voice input state, and determines that a speaker (a target sound source) is to be changed when a state in which no voice input is continued for more than a predetermined switching waiting time, and performs directivity control. The directivity control start signal is transmitted to the unit 30.

The directivity controller 30 includes a sound source area selector 40
Has a sound source area information storage unit in which the sound source area information selected in step 2 is set as a “next instruction unit”, and determines the directivity of the microphone array 11 when there is a directivity control start signal from the control timing instruction unit 20. Control is performed toward the sound source area set as the “next instruction unit”.

The sound source area selector 40 selects sound source area information associated with the keyword-sound source area comparison table 50 from the detected keywords.

The keyword-sound source region comparison table 50 is a comparison table in which the sound source specified by the detected keyword is associated with the sound source region information. The keyword-sound source area comparison table 50 is stored in a rewritable storage area in advance in consideration of the arrangement state of the target sound source (speaker).

The voice recognition / keyword detection unit 60 is a means for extracting a keyword from a conversation by word spotting, and has a voice recognition processing unit.

The keyword classification list table 70 is a list for specifying the next speaker based on the keyword detected by the voice recognition / keyword detection unit 60 or detecting the end of the utterance. The next speaker is identified.

The signal processing section of the target sound source area designating sound collection device having the above configuration operates as follows.
The external sound signal picked up from the microphone array 11 is transmitted to the voice recognition / keyword detection unit 60 via the microphone array main device 12 with only the sound of the target sound source area emphasized. Voice recognition / keyword detection unit 6
The voice signal transmitted to 0 is recognized inside the voice recognition / keyword detection unit 60, and only the keyword is cut out by word spotting.

The keywords extracted by the voice recognition / keyword detection unit 60 are classified into three types of keywords shown in Table 1 with reference to a keyword classification list table 70.

[0030]

[Table 1]

The keyword classification process will be described with reference to FIG.

The voice recognition / keyword detection unit 60 monitors whether or not there is a voice input (S1), and when there is a voice input, sets the voice input state of the control timing instructing unit 20 to "present" (S2). ). Next, the voice recognition / keyword detection unit 60 monitors the detection of the keyword (S
3). When a keyword is detected, it is determined whether or not the detected keyword is a speaker specific word such as an inventor's personal name or a nickname with reference to the keyword classification list table 70 (S4). If it is the speaker specific word, after performing the processing 1 shown in FIG. 4 (S5), it is determined whether or not the speech recognition has been completed (S6), and if it has been completed, the processing is terminated. If the voice recognition has not been completed in step S6, the process returns to step S1 to monitor the voice input.

If there is no voice input in step S1, the voice input state of the control timing instructing section 20 is set to "none" (S10), and the process returns to step S1 to monitor the voice input.

If the keyword detected in step S4 is not a speaker specific word, it is determined whether or not it is a speaker specific cancellation instruction word such as "everyone" (S11). If it is the speaker specific cancellation instruction word, the process 2 is executed (S12).
Thereafter, the process shifts to step S6 to monitor the end of the speech recognition.

If it is determined in step S11 that the instruction word is not a speaker specific cancellation instruction word, it is determined whether the instruction word is a speaker change promotion word that is a clear question sentence such as "please" or "is?" (S13). If the word is a speaker change promotion word, after the process 3 is executed (S14), the process proceeds to step S6 to monitor the end of the speech recognition.

If no keyword is detected in step S3 or if it is not a speaker change promotion word in step S13, the flow shifts to step S6 to monitor completion of speech recognition.

Processing 1 in step S5 when the keyword is a speaker specific word is processing as shown in FIG. That is, the speech recognition / keyword detecting unit 60 that determines that the detected keyword is the speaker specific word transmits the detected keyword to the sound source region selecting unit 40 (S21), and the sound source region selecting unit 40 The sound source area information corresponding to the specified speaker is selected from the area comparison table 50 (S22), and the sound source area information is set in the “next instruction section” of the directivity control unit 30 (S23). .

In the keyword-sound source area comparison table 50, the selected sound source area candidates are associated with the detected keywords, and are stored in a form that can be rewritten as needed.

In step S11, when it is determined that the keyword is the speaker specific elimination instructing word, the processing 2 is performed via the sound source area selecting section 40 and the directivity controlling section as shown in the flowchart of FIG. This is the process of resetting the "next instruction unit" in S30 (S30).

In step S13, when it is determined that the keyword is the speaker change promotion word, as shown in the flowchart of FIG. 6, the sound source area selector 40 waits for the control timing instructor 20 to switch. This is a process of setting the time to “short period” (S40).

The control timing instructing section 20 executes the processing shown in the flowchart of FIG. First, the switching waiting time of the directivity control unit 30 is set to “long time” (S
51). The elapsed time of the voice input state “None” is monitored (S
52), it is determined whether or not the voice input state is “absent” (S53). If the voice input state is "none", it is determined whether or not the elapsed time of the voice input state "none" has exceeded a set waiting time (S54). When the set waiting time is exceeded, the control timing instructing section 2
0 transmits a directivity control start signal to the directivity control unit 30 (S55), sets the switching waiting time to "long term" (S56), and determines whether the control timing instruction has been completed. A determination is made (S57).

When the control timing instruction is completed,
This processing ends.

If the voice input is "present" in step S53, the process proceeds to step S57 to determine whether or not the control timing instruction has been completed. If so, the process is terminated.

If it is determined in step S54 that the elapsed time of the voice input state "absent" does not exceed the switching waiting time, the process returns to step S53 and the presence or absence of voice input is monitored.

If the control timing instruction has not been completed in step S57, the flow returns to step S52 to monitor the elapsed time of the voice input state "none".

The switching waiting time is set to “long time” immediately after the start of the control and after the end of the directivity control, but is changed to “short time” when the processing 3 is executed.

The processing of the directivity control unit 30 will be described with reference to the flowchart of FIG. The directivity control unit 30 resets the data of the “next instruction unit” already registered prior to the directivity control process (S61), and then receives the directivity control start signal from the control timing instruction unit 20. It monitors whether or not it is, and continues monitoring until it is received (S62). When the directivity control start signal is received, it is determined whether or not the “next instruction section” data is set (S63). When the “next instruction unit” is set when the directivity control start signal is received from the control timing instruction unit 20,
A control signal for controlling the directivity of the microphone is transmitted to the microphone array main device 12 toward the instruction unit (S6).
4). If it is determined in step S63 that the “next instruction unit” is not set, a control signal for canceling the directivity control is transmitted to the microphone array main device 12 (S65).

Thereafter, it is determined whether or not the directivity control is to be terminated (S66). When the process is to be terminated, this process is terminated. When not to be terminated, the process returns to step S61 to continue the directivity control process.

In the above embodiment, more appropriate directivity control can be performed by making the following changes.

First, a keyword-sound source area comparison table 50
However, the accuracy of sound source estimation can be improved by associating the sound source region with a logical operation expression combining a plurality of keywords, instead of a simple table in which one keyword is associated with one sound source region.

Second, in the above embodiment, the directivity control unit 3
The sound source area information of 0 is rewritten to the newest one.
In the sound source area selection unit 40, priorities are assigned to keywords, and the sound source area information is transmitted only when the priority is higher than that previously transmitted, so that the accuracy of sound source estimation can be improved.

Third, in the embodiment described above, the change of the speaker is performed by obtaining a keyword from the content of the previous speaker. On the other hand, in a meeting or the like, there is a case where a facilitator of the agenda exists separately from a speaker, and the change of the speaker is forced. The target sound source area designating sound collection method in the case where the facilitator is present will be described with reference to FIG.

FIG. 2 is a block diagram showing a configuration example of a signal processing unit of a target sound source area designating sound collecting apparatus by keyword detection according to a second embodiment of the present invention. The signal processing unit of the target sound source area designating sound pickup device by keyword detection according to the second embodiment has a configuration of a signal processing unit of the target sound source region designating sound collection device by keyword detection according to the first embodiment. Priority microphone 80 exclusively for the facilitator
This is characterized in that a keyword input unit as described above and a voice selection switch 90 are provided.

In this embodiment, as the voice input to the voice recognition / keyword detection unit 60, one of the voice from the sound collection device 10 and the voice from the priority microphone 80 dedicated to the chairman is selected. Sound selection switch 90
Is entered via

The voice selection switch 90 is used when the priority microphone 80 dedicated to the chairman inputs a voice.
By inputting this voice to the voice recognition / keyword detection unit 60 prior to other keyword input (sound collecting device 10), the accuracy of sound source estimation can be improved.

[0056]

As described above, the target sound source area designating sound collecting device by keyword detection according to the present invention is a sound collecting device which specifies a sound source from a keyword and performs directivity control without performing sound collecting scanning. In addition, only the target sound can be collected at a high SN ratio, which is particularly effective in a noisy environment.

By incorporating a voice recognition unit for recognizing collected voice signals as keyword input means, directivity control can be automatically performed in real time without bothering input operations.

Further, by linking with a photographing device such as a camera, it is possible to automatically follow up and photograph a sound source. When combined with another device, it is expected that various functions can be realized.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a signal processing unit of a target sound source area designating sound pickup device by keyword detection according to a first embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of a signal processing unit of a target sound source area designating sound collection device by keyword detection according to a second embodiment of the present invention.

FIG. 3 is a flowchart for explaining the operation of a voice recognition / keyword detection unit of the device of the present invention.

FIG. 4 is a flowchart that embodies the contents of processing 1 when the keyword in the flowchart of FIG. 3 is a speaker specific word;

FIG. 5 is a flowchart that embodies the content of processing 2 when the keyword in the flowchart of FIG. 3 is a speaker identification cancellation instruction word;

FIG. 6 is a flowchart that embodies the content of processing 3 when the keyword in the flowchart of FIG. 3 is a speaker change promotion word;

FIG. 7 is a flowchart for explaining the operation of a control timing instruction unit of the device of the present invention.

FIG. 8 is a flowchart for explaining the operation of the directivity control unit of the device of the present invention.

[Explanation of symbols]

 Reference Signs List 10 sound pickup unit 11 microphone array 12 microphone array main unit 20 control timing instruction unit 30 directivity control unit 40 sound source region selection unit 50 keyword-sound source region comparison table 60 voice recognition / keyword detection unit 70 keyword classification list table 80 priority microphone 90 Voice selection switch

Claims (6)

[Claims] 1. A sound pickup device comprising: a sound pickup unit having directivity; a unit for controlling the directivity of the sound pickup unit; a unit for selecting a target sound source area for sound pickup; and a keyword input unit. The keyword and the sound source area are associated in advance, and when the keyword is input from the keyword input unit, the associated sound source area is selected, and the directivity of the sound pickup unit is adjusted according to the selected area. A sound pickup device characterized by controlling the following. 2. The keyword input device according to claim 1, further comprising: means for recognizing a voice signal obtained from the sound collecting section to obtain voice information; and means for detecting a keyword from the voice information. A sound pickup device as described. 3. The sound pickup device according to claim 1, wherein the keyword input from the keyword input means is an identifier for identifying a sound source. 4. An audio signal obtained from the sound pickup section,
2. The sound pickup according to claim 1, further comprising means for detecting a state of no sound, wherein the directivity control of the sound pickup unit is performed in synchronization with the case where the soundless state continues for a predetermined time or more. apparatus. 5. A sound pickup device in which the directivity of a sound pickup unit having directivity is controlled toward a target sound source area,
Keyword detection means for detecting a keyword specifying the next target sound source from the voice input from the sound pickup unit, a keyword classification list storing data indicating a change of the target sound source, and data in which the target sound source and the sound source region are previously associated A keyword-sound source region comparison table, a sound source region selection unit that selects a sound source region by referring to the keyword-sound source region comparison table using data on the target sound source detected by the keyword detection unit, and a directivity of the sound collection unit. Control timing instructing means for outputting a control start instruction signal for instructing the timing to start the control to shift to the selected sound source region,
A sound pickup device comprising: a directivity control unit that controls the directivity of a sound pickup unit according to a sound source area selection signal and a control start instruction signal. 6. The control timing instructing means includes means for detecting a state of no sound from a sound signal obtained from a sound pickup unit, and controls the control in synchronization with the case where the state of no sound continues for a predetermined time or more. The sound collection device according to claim 5, wherein a start instruction signal is output.