JP2002314637A - Device for reducing noise - Google Patents

Abstract

PROBLEM TO BE SOLVED: To clarify inputted voice by appropriately processing various kinds of noise that occur in a moving body. SOLUTION: This noise reducing device is provided with a factor information detecting part 28, and noise data of noise that occurs in a vehicle are stored in a storing part 2 while made to correspond to factor information detected by the factor information detecting part 28. An update selecting part 29 selects noise data corresponding to the detected factor information, and a noise calculating part 30 calculates an estimated noise spectrum on the basis of the selected noise data. An operating part 32 reduces noise by using the estimated noise spectrum. A noise estimating part 38 estimates inputted noise data from a sound signal. An update deciding part 33 decides whether or not the inputted noise data are similar to the estimated noise data, eliminates noise by using an input noise spectrum when they are not similar, and also updates the noise data in the storing part 2.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a noise reduction apparatus using spectral subtraction.

[0002]

2. Description of the Related Art With the rapid spread of portable communication terminals in recent years, the number of people who use portable communication terminals in vehicles such as automobiles is increasing. For this reason, when used in a vehicle, laws and regulations have made it mandatory to use a hands-free device or the like in order to avoid one-handed operation (driving) as when using a conventional handset. In this hands-free device, for example, a microphone for transmission is installed in a sun visor unit, and an on / off switch for transmission and reception is installed in a steering wheel unit. The audio output is configured such that a dedicated speaker or a speaker for vehicle-mounted audio is also used. Therefore, the driver can talk without releasing the hand from the steering wheel.

Further, with the advancement of a driving support system mounted on a vehicle such as an automobile, a voice recognition device for controlling various devices by voice input has been remarkably spread, and a voice input microphone provided in a vehicle compartment has been used. It performs voice recognition and controls various devices. This voice recognition device also has a configuration in which, for example, a voice input microphone is provided in a sun visor unit, and a voice recognition on / off switch (voice input-switch) is provided in a steering wheel unit. Therefore, the driver can control various devices using the voice recognition device without releasing the hand from the steering wheel.

[0004] The reduction of noise is indispensable for such voice communication and control techniques. Conventionally, when a portable communication terminal is used in a vehicle such as an automobile using a hands-free device or the like, various background noises in the vehicle (wind noise, engine sound, air conditioner motor sound, blowing sound, road noise, audio sound, exhaust sound, etc.). Sound, etc.) is input from the microphone at the same time as the call voice, and there is a problem that the quality of the transmitted voice is significantly reduced. Similarly, in the speech recognition device, there is a problem in that various background noises in the vehicle are mixed into the input speech, thereby reducing the speech recognition rate.

[0005] One of the techniques for reducing such noise is a noise canceling method for removing a noise component of speech input from a microphone. A noise spectrum is obtained from a signal in an unvoiced state, and a noise spectrum is obtained from a speech spectrum during speech. A technique for subtracting a noise spectrum at a constant rate is known. This is called spectral subtraction,
For example, STEVEN F BOLL, “Suppression of Acoustic Noi
se in Speech Using Spectral Subtruction ”, IEEE Tr
ansactions on Acoustics, Speech and Signal process
Many research results have been published, including ing, Vol. Assp-27, No. 2, April 1979, pp. 113-120.

Here, a method of fixing the magnification to be subtracted and the spectrum shape to experimentally predetermined values is generally used, but it has also been proposed to change the magnification or the spectrum shape in accordance with the power or shape of the noise. Conventionally, the following methods have been proposed as techniques using such spectral subtraction.

As shown in Japanese Patent Application Laid-Open No. 57-122498, two microphones for voice and noise are provided, input gains of the voice microphone and the noise microphone are controlled, and input to a voice recognition unit is performed. A method for keeping the noise level in the medium constant As shown in Japanese Patent Application Laid-Open No. 11-231897,
A voice section in which low-frequency noise is removed from a voice signal containing a noise signal is detected, and a noise signal spectrum immediately before the voice section is subtracted from the entire voice section from the signal obtained by converting the voice signal into a spectrum, and the voice from which the noise is removed is obtained. As shown in Japanese Patent Application Laid-Open No. H10-177394,
Method for changing the magnification of a noise pattern to be subtracted from an input speech pattern based on noise spectrum and power

[0008]

However, the techniques described in the above publications have the following problems, respectively. This is shown corresponding to the above. Problem: In this case, gain control is performed even for sudden noise (for example, road joints or steps). In many cases, the gain control for such sudden noise is not appropriate. In this case, the noise reduction processing may cause a sense of incongruity in the auditory sense or erroneous recognition.

More specifically, in the process for reducing noise, a control amount for canceling noise is calculated every moment.
However, with respect to sudden noise, the noise component often changes greatly at the time when the control amount is calculated. That is, the time delay of the control amount calculation processing has an adverse effect on the reduction of sudden noise.
Due to the improper noise reduction processing caused by such a time delay, for example, the voice arriving via the above-described hands-free device may give a sense of incompatibility to the other party. For the same reason, erroneous recognition also occurs in the speech recognition device.

In this case, the noise is calculated and removed by the noise pattern estimated immediately before the voice section, so that the noise changes during the voice section or the timing at which the noise is estimated is accidental noise. If so, a sense of incongruity may occur in the sense of hearing, or erroneous recognition may occur.

In this case, the average spectrum of noise is calculated using the input speech spectrum and the noise section of several frames before the speech section,
Subtract the noise spectrum from the input speech spectrum. Therefore, in the process of estimating the noise pattern, the same problem as in the above-described method occurs even though the accuracy of the noise pattern estimation based on the noise spectrum and noise power is improved. In particular, here, the noise spectrum cannot be extracted unless the voiced / unvoiced determination is correctly performed.

SUMMARY OF THE INVENTION The present invention has been made in view of such conventional problems, and has as its object to appropriately process various noises generated in a moving body to clarify an input voice.

[0013]

Means for Solving the Problems and Effects of the Invention The noise reduction apparatus of the present invention is used by being mounted on a moving body. The moving objects include not only cars for transporting so-called passengers and cargo such as trains, trains and cars, but also racing cars and motorcycles. In addition, not only cars but also ships and aircraft are included. This is because the purpose is to reduce noise due to movement.

In order to reduce the noise, according to the present invention, the noise data of the noise that can be estimated to be generated for the sound generated in the moving object is stored in the storage means in association with predetermined factor information. . The factor information is information that causes noise generation. Then, a configuration is provided with factor information acquiring means for acquiring the factor information, and based on the acquired factor information, the selecting means selects corresponding noise data from the noise data stored in the storage means.

Further, the noise calculating means calculates estimated noise data including an estimated noise spectrum which is a frequency spectrum of the estimated noise based on the selected noise data. On the other hand, the conversion unit acquires the audio signal input from the microphone, performs digital conversion, and the spectrum calculation unit calculates the frequency spectrum of the converted audio signal.

The calculating means subtracts the estimated noise spectrum calculated by the noise calculating means from the frequency spectrum calculated by the spectrum calculating means. Subtracting the estimated noise spectrum from the frequency spectrum may be a subtraction process of an amplitude spectrum that is an amplitude component of the frequency spectrum and the estimated noise spectrum. Further, a subtraction process of a power spectrum obtained by squaring the amplitude spectrum may be performed.

If the conventional problem is simply repeated here, noise reduction processing is performed on all noises generated in the moving object. That is, there has been a problem that noise is actually picked up by a microphone or the like, and the noise is uniformly processed.

According to the present invention, there is a case that the input speech may be clarified if it is not processed depending on the noise, and the noise which is estimated to be generated in the moving object when the moving object is limited is limited. It is a thing that paid attention to becoming something. For example, noises in automobiles include road noise, engine noise, air-conditioner ventilation noise,
A typical example is a car audio sound.

That is, information on factors that cause noise generation is obtained, an estimated noise spectrum is calculated based on noise data stored in advance corresponding to this factor information, and noise reduction is performed using the estimated noise spectrum. . As a result, only typical noise can be set as a processing target, and appropriate noise reduction processing can be realized by intentionally removing unexpected unexpected noise from the processing target. That is, various noises generated in the moving object can be appropriately processed, and the input voice can be clarified.

Incidentally, road noise and the like depend on the road surface of the vehicle. Therefore, as described in claim 2, it is conceivable to include the position information of the moving body in the factor information. This position information can be set as information on the area on the map. In this way, noise data according to the area is selected, and appropriate noise processing for each area can be performed. In this case, it is conceivable that the factor information obtaining means is realized as a configuration including a GPS receiver. Further, a configuration including a navigation system having such a position specifying function may be adopted. Further, the position information may be obtained from an external navigation system or the like without calculating the position information. In addition, when position information is acquired from the outside, a configuration in which position information specified outside the moving object is acquired using, for example, a communication terminal or a wireless device may be considered.

When the positional information is included in the factor information as described above, the present invention further comprises a traveling direction acquiring means for acquiring the traveling direction of the moving body, wherein the selecting means comprises a traveling direction acquiring means. The noise data may be selected from the storage means in consideration of the traveling direction acquired in the step (1). In this way, noise data corresponding to the next traveling area can be selected in advance, which is advantageous for real-time noise processing. For example, by storing noise data relating to noise generated due to a step on a road surface in association with an area, and selecting the noise data in advance, it is possible to process sudden noise as non-sudden noise.

In addition, as described in claim 4, it is conceivable that the factor information includes internal information indicating the state and situation of the moving object, and external information obtained from the outside. As the internal information, the information of the engine related to the engine sound, the information of the air conditioner related to the blow sound of the air conditioner, the audio information related to the car audio sound, and the like are considered. Further, the open / closed state of the window, the number of occupants, the running conditions, and the like can be mentioned. Therefore, when internal information is included in the factor information, the factor information obtaining means includes a sensor for detecting engine rotation, a torque sensor for detecting output torque of the engine, an air volume sensor for detecting the air volume and the air direction of the air conditioner, and a ventilation direction. Sensors, media information sensors and media source information sensors that detect media information and media source information of in-vehicle audio, window opening and closing sensors that detect window open / closed status, occupant sensing sensors that detect the number of occupants and occupant positions, and rainfall around the vehicle It is conceivable to use at least one of a raindrop sensor for detecting, a wiper sensor for detecting the movement of the wiper based on rainfall around the vehicle, and a fog lamp sensor for detecting lighting of a fog lamp for securing visibility by rainfall around the vehicle. Can be

On the other hand, examples of the external information include weather-related information such as a surrounding traffic volume and a strong wind. Therefore, when external information is included in the factor information, the factor information obtaining means may be configured to include at least one of a communication terminal or a wireless device that is mounted on a mobile body and can perform wireless communication. A configuration for acquiring information transmitted from a service center or the like may be considered.

The present invention is characterized in that noise data is stored in association with factor information. However, there are known noise data and noise data that can be predicted at a certain level. For example, the engine sound described above,
The air-conditioner blowing sound, car audio sound, and the like correspond to the former, while road noise and the like correspond to the latter.

For this reason, there is a possibility that the noise data stored in advance in the storage means may be significantly different from the actual noise data. Also,
Although noise generated in an automobile or the like is finite, it is also a difficult task to prepare noise data in advance for all the factor information, especially when position information is included in the factor information.

Therefore, it is desirable to adopt the configuration described in claim 5. In this case, the noise estimation unit estimates input noise data including an input noise spectrum which is a frequency spectrum of a noise component included in the acoustic signal from the conversion unit. Then, the comparing unit compares the estimated input noise data with the estimated noise data calculated by the above-described noise calculating unit. When the instruction unit issues an update instruction based on the comparison result, the update unit stores noise data based on the input noise data in the storage unit.

That is, the input noise estimated from the acoustic signal is compared with the estimated noise based on the noise data stored in the storage means. The noise data is updated. As a result, the correspondence between the predetermined factor information and the noise data becomes appropriate, and the effect of the present invention that various noises can be appropriately processed and the input voice can be clarified is remarkable.

In the configuration having such a noise data updating function, the above-mentioned instruction means issues an update instruction and sends an input noise spectrum instead of the estimated noise spectrum to the arithmetic means. Preferably, the input noise spectrum is subtracted from the frequency spectrum by the calculating means. In this way, processing is performed with more appropriate data.

Here, the comparison between the estimated noise data and the input noise data by the comparing means will be considered. As described above, since the estimated noise data and the input noise data include a frequency spectrum, comparison may be made using this frequency spectrum. However, there are bands that are easily heard by humans as noise and bands that are not. In other words, there may be a case where it is difficult to make an appropriate comparison determination as to whether or not to update only by comparison based on the spectrum shape.

Therefore, it is preferable that the estimated noise data and the input noise data include a frequency spectrum and a change amount of the frequency spectrum.
The change amount of the frequency spectrum here is the change amount of the shape itself when the spectrum shape is divided by frequency band, or the change amount of each band.

In this way, the amount of change in the spectrum is also a comparison factor. Therefore, for example, noise in a band that can be easily heard by humans is updated even if the amount of change is small, while noise in a band that is hard to hear is updated. Can be compared when the change amount becomes relatively large. As a result, an appropriate comparison by the comparing means is realized.

Similarly, if the power of the frequency spectrum or the power and the amount of change in the power are used as comparison factors, more appropriate comparison judgment can be realized. Therefore, as described in claim 8, it is conceivable that the estimated noise data and the input noise data include, in addition to the frequency spectrum, the power of the frequency spectrum or the power and the amount of change in the power. The power here may be a power spectrum obtained by squaring the amplitude spectrum when the frequency spectrum is an amplitude component having an amplitude component. When the frequency spectrum is a power spectrum, sound energy obtained by integrating the power spectrum on the frequency axis may be used. This is simply energy without the concept of frequency. If the comparison judgment is performed including the power as described above, the comparison judgment by the comparing means becomes more appropriate.

In the present invention, in accordance with the factor information,
As described above, noise data is stored in the storage unit. At this time, claim 9
As shown in (1), noise data may be stored as noise-specific data for each noise type. This is based on the premise that if the number of moving objects is limited, noise generated is limited. In this case, when a plurality of noise data are selected by the selection unit, the noise calculation unit combines the noise data to calculate estimated noise data including an estimated noise spectrum. Thus, preparing noise data as noise-specific data is advantageous in that addition / update of noise data becomes easy, for example, when adding new noise.

[0034] Further, it is preferable that noise data is stored by coding. This is because storing in the form of a spectrum itself requires a huge storage capacity. This can contribute to a reduction in the storage capacity of the storage unit. In the case where the noise data is encoded and stored, the configuration having the function of updating the noise data may further include a code data generating means, and the code data generating means may include an input noise. Based on the spectrum, encoded noise data is generated. Specifically, as set forth in claim 12, the code data generating means selects an appropriate frequency spectrum from the frequency spectrum of noise prepared in advance so as to be closest to the input noise spectrum in terms of audibility,
The selected frequency spectrum is coded to generate noise data. The frequency spectrum of the noise prepared in advance is a noise spectrum indicating noise for each band, a noise spectrum weighted to the noise spectrum for each band, or a noise spectrum for only the low band. The noise spectrum may be constituted by at least one of the following noise spectra. Weighting is performed for each band, as described above, because there are bands that are easy for humans to hear and bands that are not. The reason why only the noise spectrum of only the low band may be prepared is that noise in the vehicle is concentrated in the low band. By doing so, even in a configuration having a function of updating noise data, coded noise data can be stored, and the storage capacity can be reduced.

Further, from the viewpoint of reducing the storage capacity, a configuration is adopted in which the interpolation means interpolates the estimated noise spectrum based on the factor information obtained by the factor information obtaining means. You may. Specifically, it is conceivable to perform noise gain interpolation according to the moving speed as described in claim 15 or to perform frequency interpolation of noise according to the engine speed as described in claim 16. Can be This technical idea is to reduce the underlying noise data by interpolating the estimated noise spectrum. Even with such a configuration, the storage capacity can be reduced.

[0036]

Embodiments of the present invention will be described below with reference to the drawings. [First Embodiment] FIG. 1 shows the configuration of a sound processing apparatus according to a first embodiment which embodies a "noise reduction apparatus" of the present invention.

The sound processing apparatus of the first embodiment includes a noise reduction circuit 1, a microphone 3, and an output unit 4 as main parts. More specifically, the noise reduction circuit 1 includes a storage unit 2,
It comprises a factor information detection unit 28, a selection unit 29, a noise calculation unit 30, a time-frequency conversion unit 31, and a calculation unit 32.

The microphone 3 inputs an acoustic signal in the vehicle. The analog / digital converter 5 converts an audio signal input from the microphone 3 into a digital signal. The output unit 4 outputs an acoustic signal from which noise components have been removed by the noise reduction circuit 1. The output form depends on the connected application.
A time-converted sound can be considered. When the output unit 4 is configured by a speaker, the output unit 4 may be configured by a dedicated speaker, or may be shared by a speaker of an in-vehicle audio. The output destination is not limited, and can be connected to various application devices having a voice interface, and can be used for, for example, voice recognition and portable terminals.

Next, each configuration of the noise reduction circuit 1 will be described. The storage unit 2 stores noise-specific spectra for each type of noise generated in the vehicle as noise data in association with the factor information detected by the factor information detection unit 28. The factor information is information that causes noise generation, and what kind of information is detected will be described later.

The selecting section 29 selects, from the storage section 2, noise data corresponding to the factor information detected by the factor information detecting section 28. The factor information detection unit 28, as shown in FIG.
A position detection unit 27 for detecting a current position, a vehicle information detection unit 37 for detecting internal information indicating the state and situation of the vehicle, and an external information detection unit 3 for collecting external information provided from outside
6 is provided.

As shown in FIG. 3, the position detector 27 includes a GPS receiver 40 for specifying a current position based on a signal from an artificial satellite and a navigation system 2 having map information.
6. a gyro sensor 41 for detecting the amount of change in the direction of the vehicle;
The vehicle includes a vehicle speed sensor 12 for detecting a traveling speed of the vehicle, an acceleration (G) sensor 11 for detecting an acceleration of the vehicle, and an image sensor 42 for capturing a situation around the vehicle as an image. As a result, position information and map information such as the current position / traveling direction and peripheral road shapes (road width, lanes, intersections, curves, bridges, tunnels, etc.) can be obtained.

As shown in FIG. 4, the vehicle information detecting section 37 acquires internal information indicating the state and situation of the vehicle, which can be acquired inside the vehicle. FIG. 4 shows the types of information to be obtained. Specifically, the engine speed 10, the engine torque 6, the air conditioner information 13, the audio information 14,
The window opening / closing state 15, the number of occupants 18, the occupant position 19, the wiper information 43, the fog lamp information 44, and the raindrop information 45 are acquired. Here, the engine speed 10 is the engine speed detected by the engine rotation sensor, the engine torque 6 is the engine torque detected by the torque sensor, and the air conditioner information 1
Reference numeral 3 denotes an air-conditioning air flow detected by the air-conditioner control unit and a direction of air flow.
D, media information such as cassettes and media source information, window opening / closing state 15 is information on the window opening / closing state of the vehicle detected by the window control unit, and the number of occupants 18 and the occupant position 19 are pressure sensors and the Information on the number of occupants and the occupant position in the vehicle detected by an infrared sensor or the like provided around the seat, wiper information 43 is on / off information based on a wiper switch, and fog light information 44 is on / off based on a fog light switch. The information, raindrop information 45 is information on the presence or absence of raindrops by a raindrop sensor. The wiper information 43, the fog lamp information 44, and the raindrop information 45 are weather-related information that causes noise.

As shown in FIG. 5, the external information detector 36 is configured to collect external information provided from outside the vehicle from a portable communication terminal or a communication device in the vehicle while the vehicle is running. FIG. 5 also shows information to be collected, similarly to FIG. Specifically, information of traffic information 21, weather 22, date and time 23, wind direction 24, and temperature 25 is acquired. The traffic information 21 is road traffic information of the current traveling area and traveling direction available from the VICS system, the road traffic information center and the Internet, and the weather 22 is the current traveling area and traveling direction available from the road traffic information center and the Internet. The weather information, date and time 23 are the current date and time available from a time signal and the Internet, the wind direction 24 is the wind direction information on the current running and traveling directions available from the road traffic information center and the Internet, etc., and the temperature 25 is the road traffic information center and the Internet etc. This is the temperature information of the vicinity of the current traveling and the traveling direction, which can be obtained from.

The noise calculating section 30 in FIG. 1 calculates estimated noise data based on the noise data selected by the selecting section 29. The noise data is a noise-specific spectrum as described above. Therefore, estimated noise data is calculated as an estimated noise spectrum obtained by combining the selected noise-specific spectra. The calculated estimated noise spectrum is output to the operation unit 32.

On the other hand, the time-frequency converter 31 calculates a frequency spectrum from the audio signal from the analog / digital converter 5 and outputs the frequency spectrum to the calculator 32. Arithmetic unit 32
Performs noise removal by subtracting the estimated noise spectrum calculated by the noise calculator 30 from the frequency spectrum from the time-frequency converter 31. Then, the output from the arithmetic unit 32 is the output from the noise reduction circuit 1.

The configuration of the sound processing apparatus according to the first embodiment and the function of each configuration have been described above. Here, the operation of the sound processing apparatus of the first embodiment will be described with reference to FIG. 1 again.
The acoustic signal in the vehicle detected by the microphone 3 is
The signal is input to the analog / digital converter 5 and is converted into a digital signal, which is converted into a frequency spectrum by the time-frequency converter 31.

On the other hand, an estimated noise spectrum is calculated by a noise calculating section 30 based on the noise data selected by the selecting section 29 based on the factor information output from the factor information detecting section 28. Then, the arithmetic unit 32 performs noise reduction by subtracting the estimated noise spectrum from the frequency spectrum.

At this time, it is conceivable to realize noise reduction more effectively by removing a known sound source included in the acoustic signal in the vehicle in advance. FIG. 6 shows the configuration of an input signal correction unit as a configuration example for removing a known sound source included in an audio signal. As shown in FIG. 6, when the audio output is on based on the audio information 14, an input signal obtained by convolving the interior reverberation characteristic (impulse response) with the audio media source is obtained, and this signal is time-frequency-converted. A higher noise reduction effect can be obtained by converting the signal into a signal having a frequency spectrum by the converter and subtracting the signal from the input audio signal into the spectrum of the audio signal.

In this embodiment, the noise data stored in the storage unit 2 is a noise-specific spectrum for each noise type. In this case, it is advantageous in that addition / update of the noise data becomes easy. However, instead of the spectrum for each noise, the spectrum of the noise itself (noise spectrum) corresponding to the combination of the factor information may be used.

Next, the effects of the sound processing apparatus of the first embodiment will be described. Here, in order to facilitate understanding of the description here, conventional problems will be described first. Conventionally, noise was actually picked up by a microphone or the like, and the noise was uniformly processed. for that reason,
It performs noise reduction processing even for sudden noise, which in turn causes auditory discomfort,
This resulted in misrecognition in speech recognition.

That is, depending on the noise, it may be possible to clarify the input voice if it is not processed. In addition, typical noise that is estimated to occur in a vehicle is limited. Therefore, in the present embodiment, a configuration is provided in which the factor information detection unit 28 is provided, and noise data of noise estimated to be generated in the vehicle is previously stored in the storage unit 2 in accordance with the factor information detected here. Keep it. Then, the selection unit 29 selects noise data corresponding to the detected factor information, and the noise calculation unit 30 calculates an estimated noise spectrum based on the selected noise data. The operation unit 32 performs noise reduction using the estimated noise spectrum.

As a result, only typical noise can be processed, and appropriate noise reduction processing can be realized by intentionally removing unexpected unexpected noise from the processing target.
That is, various noises generated in the vehicle can be appropriately processed, and the input voice can be clarified.

Further, in the present embodiment, the position information of the vehicle, the information indicating the state / situation of the vehicle, and the information obtained from the outside are included in the factor information which is a factor of the noise. The detection is performed by the detection unit 37 and the external information detection unit 36. As a result, noise data is prepared in the storage unit 2 for most of the noise estimated to be generated in the vehicle, and appropriate noise removal is performed.

At this time, since the position detecting section 27 can also detect the traveling direction of the vehicle, the selecting section 29 takes the traveling direction into account and reads out the noise data of the next traveling area in advance. It may be. This makes it possible to cover a delay in the processing of the arithmetic unit 32. For example, by storing data of noise generated due to a step on a road surface and selecting the noise data in advance, sudden noise can be suddenly reduced. Can also be processed as not.

Specifically, it is conceivable to predict the sudden noise by setting the following areas. Since the position detection unit 27 includes the navigation system 26, map information is detected in addition to the position information. Therefore, as shown in FIG. 7, set the area on the map,
It is conceivable to store noise data corresponding to this area. FIG. 7 shows a state in which areas are set at predetermined intervals on a road, and noise data having various characteristics is stored in correspondence with the areas. For example, area 10
No. 3 shows that noise data having the characteristic B is stored.

The vehicle 100 shown in FIG.
It is assumed that the vehicle is traveling in the direction of. Since the noise data at the current location has the feature D, an estimated noise spectrum having the feature D is selected and output. At this time, since the traveling direction 101 and the current position are known, the traveling direction 1
It is possible to predict that the noise spectrum changes from the feature D to the features C and B in the direction of 01. Here, sudden noise source 1
If the position of 02 is predicted, the processing of the arithmetic unit 32 is not delayed, and sudden noise can be processed.

In the example shown here, the transition of the noise data is predicted in consideration of the traveling direction 101 of the vehicle 100. However, in addition to the traveling direction 101, the vehicle speed, the acceleration, and the engine speed 10 are also considered. If this is the case, the prediction accuracy of the noise data, especially the sudden noise source 102, is further improved. Further, in this example, the noise data has been described as a total of five types of A to D including the sudden noise source, but it is needless to say that the present invention is not limited to this.

In the present embodiment, the analog / digital converter 5 corresponds to “conversion means”, the time-frequency conversion section 31 corresponds to “spectrum calculation means”, and the storage section 2 corresponds to “storage means”. Equivalent to. Also, the factor information detection unit 2
8 corresponds to a “factor information acquiring unit”, the selecting unit 29 corresponds to a “selecting unit”, and the noise calculating unit 30 corresponds to a “noise calculating unit”.
And the calculation unit 32 corresponds to “calculation means”. Here, the position detection unit 27 of the factor information detection unit 28 corresponds to “traveling direction acquisition means”. [Second Embodiment] In the first embodiment, the noise data is stored in the storage unit 2 in advance in association with the factor information.

However, the noise data includes the known data,
Some are predictable at some level. For example, engine sound, air-conditioner blowing sound, car audio sound, and the like correspond to the former, while road noise and the like correspond to the latter. Therefore, the noise data stored in the storage unit 2 in advance may be significantly different from the actual noise data. Although noise generated in the vehicle is finite, especially when the position (area) of the vehicle is included in the factor information as in the first embodiment, noise data is prepared in advance for all the factor information. It can be a daunting task.

In the second embodiment, the noise data in the storage section 2 is stored and updated based on actual noise. FIG. 8 is a block diagram illustrating the configuration of the sound processing apparatus according to the second embodiment. This acoustic processing device includes a noise reduction circuit 1, a microphone 3, and an output unit 4 as main components. Further, the noise reduction circuit 1 includes a storage unit 2, a factor information detection unit 28, an update selection unit 29, a noise calculation unit 30, a time-frequency conversion unit 31, a calculation unit 32, an update determination unit 33, a noise estimation unit 38, a code It is comprised by the conversion part 46.

Here, the configurations of the noise reduction circuit 1 and the noise calculator 30 are partially different from those of the first embodiment.
Further, an update selection unit 29 is used instead of the selection unit 29 in FIG. However, in order to avoid complication, the same reference numerals are given. Note that the configurations of the storage unit 2, the factor information detection unit 28, the time-frequency conversion unit 31, and the calculation unit 32 are the same as those in the first embodiment, and a description thereof will be omitted.

In the second embodiment, in addition to the selection of the noise data from the storage unit 2, the update selection unit 29
The coded noise data output from the
To update the noise data. The noise calculation unit 30
In the second embodiment, as shown in FIG. 9, a noise spectrum calculator 34 that calculates an estimated noise spectrum based on noise data, and a noise power is calculated based on the estimated noise spectrum calculated by the noise spectrum calculator 34. Power calculation unit 3
5 is comprised. Therefore, the estimated noise data consists of the estimated noise spectrum and its noise power.
The power here may be a power spectrum obtained by squaring the amplitude spectrum, or may be sound energy integrated on the frequency axis. In addition, since the coded noise data is output from the update selection unit 29, the noise spectrum calculation unit 34 calculates an estimated noise spectrum based on the coded noise data.

The noise estimation section 38 estimates input noise data, which is a noise component included in the audio signal from the analog / digital conversion section 5. The input noise data includes an input noise spectrum which is a frequency spectrum of the input noise and a noise power of the input noise spectrum. For the noise estimation unit 38, a noise removal algorithm generally used conventionally can be used. For example, a gain adjustment method, a noise spectrum estimation method, a Kalman filter method, or the like can be used. In this embodiment, as an example, a noise spectrum of a non-voice section is calculated using time-frequency conversion, and Is removed. At this time, it goes without saying that the sound signal can be corrected in the same manner as in the first embodiment.

The update determining section 33 compares the estimated noise data calculated by the noise calculating section 30 with the input noise data calculated by the noise estimating section 38. Then, based on the result of the comparison, an update instruction is issued to the update selection unit 29 as necessary, and a switch is switched so that the input noise spectrum becomes a calculation target of the calculation unit 32. That is, depending on the determination result by the update determination unit 33, the input noise spectrum is replaced with the input noise spectrum instead of the estimated noise spectrum.
2 is input.

The coding section 46 codes the input noise spectrum in the input noise data and outputs it to the update selecting section 29. Specifically, the noise data stored by the update selection unit 29 is coded identification information (ID). As shown in FIG. 10, the coding unit 46 calculates a combination that is closest to the input noise spectrum in terms of audibility from all combinations of the noise source and the gain source prepared in advance. At this time, in consideration of the fact that the noise characteristics in the vehicle are concentrated on the low-frequency side, the noise source may be a sound source that has been subjected to sub-banding or frequency division with weighting on the low-frequency side. This makes it possible to increase the noise reduction effect with respect to the number of sound sources.

Next, the operation of the sound processing apparatus will be described with reference to the flowchart shown in FIG. First, in a first step (hereinafter, steps are simply indicated by a symbol S) 51, a digital signal detected by the microphone 3 and converted by the analog / digital converter 5 is input. As a result, the digital audio signal is input to the time-frequency converter 31 and the noise estimator 38.

The time-frequency converter 31 calculates the frequency spectrum of the audio signal converted by the analog / digital converter 5. At this time, as in the above embodiment, when the audio output is turned on based on the audio information 14, the audio media source is changed from the input signal to the reverberation in the vehicle as shown in FIG. May be subtracted in consideration of the above.

At S52, the vehicle information detector 37 detects various information. In the next step S53, the position information including the position of the vehicle currently traveling by the position detection unit 27, etc.
Detect map information. Further, in next S54, the information provided from outside the vehicle is detected by the external information detection unit 36. The information detected in S52 to S54 is output to the update selection unit 29.

At S55, the noise estimating unit 38 estimates input noise data, which is a noise component included in the audio signal, from the audio signal in the vehicle input at S51. The input noise data is coded by the coding unit 46 and output to the update selection unit 29, and is also output to the update determination unit 33.

In S56, on the premise that voice interfaces of various applications are connected to the apparatus, it is determined whether or not the connected voice interfaces are operating. This is a determination process necessary to realize only one microphone 3. That is, here, it is determined whether or not the microphone 3 is used for inputting a sound signal to the sound interface.

Here, it is determined whether the voice section is detected by the voice section detector, the voice input switch of the voice recognition device is turned on, or the speech switch of the hands-free device is turned on. Perform Usually, a hands-free device or a voice recognition device always presses a switch installed on a steering wheel or the like when used. Therefore, by detecting ON / OFF of the switch, it is possible to reliably detect whether or not the operation is being performed. Also, the initials operate in the same manner as when they are used regardless of the presence or absence of the hands-free or voice recognition device. If it is determined that the device is in use (S56: YES), the process proceeds to S58. On the other hand, if it is determined that it is not in use (S56: NO), S57
Move to.

At S57, the coded input noise spectrum is converted by the update selection unit 29 as noise data.
The information is stored in the storage unit 2 in association with the factor information. in this case,
New storage corresponding to the factor information is performed for the area where the vehicle has traveled for the first time, and update storage corresponding to the factor information is performed for the area that has traveled in the past.

When the voice interface is being used, the update selecting section 29 selects the noise data stored corresponding to the factor information from the storage section 2 and outputs the selected noise data to the noise calculating section 30. Thereby, the noise calculation unit 30 outputs the estimated noise data to the update determination unit 33.

Therefore, in S58, the noise calculation unit 30
Then, it is determined whether or not the estimated noise data calculated in step (1) and the input noise data calculated in the noise estimating unit 38 are similar. The update determination unit 33 performs similarity determination using the input noise spectrum, the estimated noise spectrum, and the noise power of each spectrum.

Even if the shapes of the estimated noise spectrum and the input noise spectrum are similar, if the powers of the two are different,
Even if the estimated noise spectrum and the input noise spectrum have the same power, if the two have different spectral shapes, the sound will be heard as completely different sounds by human ears. Therefore, by performing the comparison including the power of the spectrum, it is possible to make a more accurate determination. Furthermore, it is also conceivable to add the amount of change in the spectrum and the amount of change in the power to the estimated noise data and the input noise data, and use these amounts of change in the determination.

If it is determined that the two data are similar (S58: YES), the estimated noise spectrum is output to the arithmetic unit 32 in S59, and thereafter, the process proceeds to S62. On the other hand, when it is determined that the two data are not similar (S58: NO), the input noise spectrum is output to the arithmetic unit 32 in S60, and the updating process is performed in S61, and thereafter, the process proceeds to S62. . Note that S59 and S60
Is realized by the switch control by the update determination unit 33 (see FIG. 8). The update processing in S61 is the same processing as S57 described above.

In S62, the operation unit 32 removes noise. This is a spectrum of noise output from the frequency spectrum converted by the time-frequency conversion unit 31,
That is, one of the estimated noise spectrum and the input noise spectrum is subtracted. That is, in the acoustic processing device of the second embodiment, the input noise data estimated from the acoustic signal by the noise estimating unit 38 and the estimated noise calculated by the noise calculating unit 30 from the noise data stored in the storage unit 2. It is determined whether or not the data is similar (S5 in FIG. 11).
8) If similar, noise removal is performed using the estimated noise spectrum (S59, S62), while if not similar, noise removal is performed using the input noise spectrum (S6).
0, S62). Then, at this time, the noise data in the storage unit 2 is updated with the noise data based on the input noise spectrum (S61).

When the noise data in the storage unit 2 is updated in this manner, the correspondence between the factor information and the noise data becomes appropriate, so that various noises can be appropriately processed and the input voice can be clarified. The effect described in the first embodiment, which is achieved, is further enhanced. The update timing of the noise data is not limited to the above embodiment. This is because it differs depending on the application provided with the voice interface connected to the present apparatus. For example, the update timing may be a fixed time interval, a fixed travel distance interval, a time when the noise spectrum fluctuates more than a threshold value, or a time when the noise gain fluctuates more than a threshold value.

Further, in the second embodiment, as shown in FIG. 9, estimated noise data including the estimated noise spectrum and its power is calculated by the noise calculating section 30, and not only the spectrum but also the power is used. We make similar judgments. This makes it possible to make a more accurate update determination.

Further, in the second embodiment, it is determined whether or not the voice interface connected to the apparatus is in use (S56 in FIG. 11). The obtained acoustic signal is stored as a noise component as noise data corresponding to an area which is one of the position information. This makes it possible to deal with any noise in any place where you have traveled in the past. Also,
It is possible to predict a sudden noise source or a noise pattern existing in the traveling direction from the current position, and it is possible to prevent a delay in the processing by the arithmetic unit 32. False recognition can be reliably prevented.

In the second embodiment, the coding section 46
The input noise spectrum estimated by the noise estimating unit 38 is stored in the storage unit 2 as coded code data. This is because when noise data is stored in the form of the spectrum itself, each one becomes larger,
This is because the total storage amount increases in proportion to the number of storages.
That is, spectrum data focused on the noise in the vehicle compartment is prepared, and the spectral data closest to the audibility is converted into identification information (ID) and stored, so that the memory usage can be significantly reduced.

Note that the sound processing apparatus of the second embodiment also
As in the first embodiment, the analog / digital converter 5 corresponds to “converter”, the time-frequency converter 31 corresponds to “spectrum calculator”, and the storage 2 corresponds to “storage”. The factor information detecting unit 28 corresponds to “factor information obtaining unit”, the noise calculating unit 30 corresponds to “noise calculating unit”, the calculating unit 32 corresponds to “calculating unit”, and the factor information detecting unit 28 The position detection unit 27 corresponds to “traveling direction acquisition means”.

Further, the noise estimating section 38 corresponds to “noise estimating means”, and the update selecting section 29 corresponds to “selecting means” and “updating means”. Further, the update determination unit 33 corresponds to “instruction unit” and “comparison unit”, and the coding unit 46 corresponds to “coding unit”. Also, S52 to S in FIG.
The processing of 54 corresponds to the processing as factor information detecting means,
The processing in S55 corresponds to the processing as noise estimation means.
The processing of 58 to S60 corresponds to the processing as the instruction means,
The process of S61 corresponds to a process as an updating unit.

As described above, the present invention is not limited to such an embodiment, and can be implemented in various forms without departing from the gist of the present invention. For example, the following embodiment in which a part of the configuration of the first and second embodiments is changed is also conceivable. [Third Embodiment] The acoustic processing apparatus shown in FIG. 12 has a configuration in which the storage unit 2 described in the first embodiment stores the estimated noise spectrum itself as noise data corresponding to the factor information. In this case, since the read noise data is already the estimated noise spectrum, the noise calculation unit 30 becomes unnecessary. Fourth Embodiment The acoustic processing apparatus shown in FIG. 13 has a configuration in which the storage unit 2 described in the second embodiment stores the estimated noise spectrum itself as noise data corresponding to the factor information. In this case, the noise calculating unit 30 becomes unnecessary as in the third embodiment, and the input noise spectrum is stored in the storage unit 2 as it is when updating the noise data. Also becomes unnecessary. Fifth Embodiment By the way, it has already been described that the storage capacity of the storage unit 2 can be reduced by coding the noise data. Such a reduction in storage capacity can also be realized by a configuration that interpolates noise data in the storage unit 2 as shown in FIG.

In the first to fourth embodiments, the storage unit 2
Noise data is stored in correspondence with all the factor information, that is, in correspondence with every combination of the factor information. On the other hand, the storage unit 2 stores noise data thinned out according to a predetermined rule. Then, by providing the noise interpolation unit 47, the estimated noise spectrum based on the decimated noise data is interpolated. As a result, performance equivalent to that of the above embodiment can be exhibited while reducing the storage capacity of noise data.

This will be described. Based on the noise-specific spectrum selected by the selection unit 29, the noise calculation unit 30
Calculates the estimated noise spectrum (before interpolation) as in the first embodiment. However, the estimated noise spectrum calculated here has low accuracy because it is calculated based on the thinned noise data. Therefore, the noise interpolation unit 47 interpolates the estimated noise spectrum based on the factor information from the factor information detection unit 28.

For example, when traveling at a vehicle speed of 60 km / h and an engine speed of 2000 rpm, the vehicle speed
The following description is based on the assumption that an estimated noise spectrum has been calculated from noise data corresponding to 0 km / h and an engine speed of 1000 rpm. As shown in FIG. 15, the noise interpolation unit 47 includes a vehicle speed, an engine speed 10, a window opening / closing state 15,
The information such as the audio information 14 is input, and the interpolation corresponding to the current engine speed 10 is obtained from the frequency-gain characteristics corresponding to the engine speed 10 stored according to a predetermined rule as shown in FIG. The value is estimated by linear interpolation. In this example, the value is an intermediate value between 1000 rpm and 3000 rpm.

Further, an interpolation value corresponding to the current vehicle speed is estimated by linear interpolation from the frequency-gain characteristics corresponding to the vehicle speed stored according to a predetermined rule as shown in FIG. In this example, 40km / h and 80km
It is an intermediate value of m / h. By doing so, the engine speed 1 is added to the estimated noise spectrum based on the noise data.
By calculating (adding) the interpolated value of 0 and the interpolated value with respect to the vehicle speed, an estimated noise spectrum in which characteristics are interpolated can be calculated. The interpolated estimated noise spectrum has an accuracy comparable to that of the first embodiment. On the other hand,
The data amount of the noise data is reduced, and the storage capacity of the storage unit 2 can be reduced. [Others] A function of updating noise data may be added to the configuration of the fifth embodiment. An example of the configuration is as shown in FIG.

Further, an embodiment in which the characteristic portions of the first to fifth embodiments are appropriately combined may be adopted. That is, a configuration in which the interpolation function of the fifth embodiment is combined with the configuration of the third embodiment can be considered. As shown in FIG. Further, the interpolation function of the fifth embodiment may be combined with the fourth embodiment. As shown in FIG. [Specification of Vehicle Position] In the present embodiment, the position of the vehicle is specified by the position detector 27 using the GPS receiver 40 or the like. It may be specified by using.
That is, as shown by a broken line in FIG.
To have a communication terminal (or radio) 48,
It is conceivable to obtain the position information via the communication terminal 48. Further, considering that the position is specified by external information, the external information detecting unit 36 may acquire the position information of the vehicle. That is, the external information detection unit 36 may obtain the position information 49 as indicated by a broken line in FIG. At present, position detection using GPS is the mainstream, but in the future, when infrastructure devices are provided, it is conceivable that the vehicle position will be specified by these infrastructure devices.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram illustrating a configuration of a sound processing apparatus according to a first embodiment.

FIG. 2 is an explanatory diagram showing a factor information detection unit.

FIG. 3 is an explanatory diagram showing a position information detection unit.

FIG. 4 is an explanatory diagram showing a vehicle information detection unit.

FIG. 5 is an explanatory diagram showing an external information detection unit.

FIG. 6 is an explanatory diagram illustrating an example of an input signal processing unit.

FIG. 7 is an explanatory diagram showing a method for predicting a sudden noise source.

FIG. 8 is an explanatory diagram illustrating a configuration of a sound processing device according to a second embodiment.

FIG. 9 is an explanatory diagram illustrating a configuration of a noise calculator according to a second embodiment.

FIG. 10 is an explanatory diagram showing a coding method by a coding unit.

FIG. 11 is a flowchart illustrating an operation of the sound processing device of the second embodiment.

FIG. 12 is an explanatory diagram illustrating a configuration of a sound processing device according to a third embodiment.

FIG. 13 is an explanatory diagram illustrating a configuration of a sound processing apparatus according to a fourth embodiment.

FIG. 14 is an explanatory diagram illustrating a configuration of a sound processing device according to a fifth embodiment.

FIG. 15 is an explanatory diagram illustrating an outline of a noise interpolation unit.

FIG. 16 is an explanatory diagram illustrating a gain-frequency characteristic of an engine speed.

FIG. 17 is an explanatory diagram illustrating a gain-frequency characteristic of a vehicle speed.

FIG. 18 is an explanatory diagram illustrating a configuration of a sound processing device according to another embodiment.

FIG. 19 is an explanatory diagram illustrating a configuration of a sound processing device according to another embodiment.

FIG. 20 is an explanatory diagram illustrating a configuration of a sound processing device according to another embodiment.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Noise reduction circuit 2 ... Storage part 3 ... Microphone 4 ... Output part 5 ... Analog / digital conversion part 6 ... Engine torque 10 ... Engine speed 11 ... Acceleration sensor 12 ... Vehicle speed sensor 13 ... Air conditioner information 14 ... Audio information 15 ... window opening / closing state 18 ... number of occupants 19 ... occupant position 21 ... traffic information 22 ... weather 23 ... date and time 24 ... wind direction 25 ... temperature 26 ... navigation system 27 ... position detection unit 28 ... factor information detection unit 29 ... selection unit, update selection Unit 30 noise calculation unit 31 time-frequency conversion unit 32 calculation unit 33 update determination unit 34 noise spectrum calculation unit 35 power calculation unit 36 external information detection unit 37 vehicle information detection unit 38 noise estimation unit 40 GPS receiver 41 Gyro sensor 42 Image sensor 43 Wiper information 44 Fog lamp information 4 5 raindrop information 46 coding part 47 noise interpolation part 48 communication terminal 49 position information 100 vehicle 101 traveling direction 102 sudden noise source 103 area

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G10L 15/28 G10L 3/00 551J 21/02 521L H04B 1/10 531Q 551Q (72) Inventor Isao Aichi Aichi 1-chome, Showa-cho, Kariya-shi, Japan F-term in DENSO Corporation (reference) 3D020 BA07 BB01 BC01 BD14 BE03 5D015 EE05 5K027 AA11 AA16 BB03 DD12 HH03 5K052 AA01 BB08 DD02 EE12 EE24 FF31

Claims (16)

[Claims] 1. A conversion unit for acquiring an audio signal input from a microphone and being used by being mounted on a moving body and performing digital conversion, and calculating a frequency spectrum of the audio signal converted by the conversion unit. Spectrum calculating means, and storage for storing noise data, which is noise data whose generation can be estimated for a sound generated in the moving object, in association with predetermined factor information which is a factor of the noise generation. Means, factor information obtaining means for obtaining the factor information, based on the factor information obtained by the factor information obtaining means,
Selecting means for selecting corresponding noise data from the storage means; and noise calculation for calculating estimated noise data including an estimated noise spectrum which is a frequency spectrum of estimated noise based on the noise data selected by the selecting means. And a calculating means for subtracting the estimated noise spectrum calculated by the noise calculating means from the frequency spectrum calculated by the spectrum calculating means. 2. The noise reduction device according to claim 1, wherein the factor information includes position information of a moving body. 3. The noise reduction device according to claim 2, further comprising a traveling direction acquisition unit that acquires a traveling direction of the moving body, wherein the selection unit includes a traveling direction acquired by the traveling direction acquisition unit. Wherein the noise data is selected from the storage means in consideration of the following. 4. The noise reduction device according to claim 1, wherein the factor information includes at least one of internal information indicating a state / situation of the moving body or external information acquired from outside the moving body. A noise reduction device characterized by including one of them. 5. The noise reduction apparatus according to claim 1, wherein input noise data including an input noise spectrum which is a frequency spectrum of a noise component included in the acoustic signal from the conversion unit is estimated. Noise estimating means; comparing means for comparing input noise data estimated by the noise estimating means with estimated noise data calculated by the noise calculating means; and the noise data based on a comparison result by the comparing means And updating means for updating the noise data by storing the noise data based on the input noise data in the storage means when there is an updating instruction by the instruction means. Noise reduction device. 6. The noise reduction apparatus according to claim 5, wherein said instruction means issues said update instruction and sends said input noise spectrum instead of said estimated noise spectrum to said arithmetic means. Wherein the input noise spectrum is subtracted from the frequency spectrum when the input noise spectrum is transmitted by the instruction means. 7. The noise reduction apparatus according to claim 1, wherein the estimated noise data and the input noise data include a change in the frequency spectrum in addition to a frequency spectrum. Noise reduction device. 8. The noise reduction apparatus according to claim 1, wherein the estimated noise data and the input noise data are, in addition to a frequency spectrum, a power of the frequency spectrum or a power of the frequency spectrum and the power. A noise reduction device characterized by including a change amount of the noise. 9. The noise reduction device according to claim 1, wherein the noise data is stored as noise-specific data for each noise type. 10. The noise reduction device according to claim 1, wherein the noise data is coded and stored. 11. The noise reduction device according to claim 10, further comprising: generating said coded noise data based on said input noise spectrum, provided that said noise reduction device has a function of updating said noise data. A noise reduction device comprising code data generation means. 12. The noise reduction apparatus according to claim 11, wherein said code data generating means selects an appropriate frequency from a frequency spectrum of noise prepared in advance so as to be audibly closest to said input noise spectrum. A noise reduction apparatus comprising: selecting a spectrum; encoding the selected frequency spectrum to generate noise data. 13. The noise reduction device according to claim 12, wherein the frequency spectrum of the previously prepared noise is a noise spectrum indicating noise for each band, a noise spectrum weighting the noise spectrum for each band, or A noise reduction device comprising a noise spectrum of at least one of a noise spectrum of only a low band. 14. The noise reduction device according to claim 1, further comprising an interpolation unit that interpolates the estimated noise spectrum based on the factor information obtained by the factor information obtaining unit. A noise reduction device characterized by the above-mentioned. 15. The noise reduction apparatus according to claim 14, wherein said interpolation means performs noise gain interpolation according to a moving speed. 16. The noise reduction device according to claim 14, wherein said interpolation means performs frequency interpolation of noise according to an engine speed.