A method for acquiring anti-reverberation speech signals suitable for target building space

Abstract

The present invention proposes a method for acquiring an anti-reverberation speech signal suitable for a target building space. First, an impulse response signal of the target building space is collected. Then, a human-computer real-time interactive simulation system is used to allow two experimenters to perform a dialogue task. During the dialogue task, the two experimenters communicate and continuously obtain feedback, and continuously correct their own voices according to the feedback. Finally, the voices of the experimenters are stabilized and both parties can hear clearly, thereby obtaining speech data in a stable voice state that the experimenters can hear clearly; the voice characteristics of the speech data in a stable voice state that the experimenters can hear clearly are obtained by analyzing and learning, and an anti-reverberation speech signal suitable for the target building space is obtained; the generated anti-reverberation speech signal is input into the broadcasting system of the target building space for playback, so as to achieve a good voice communication effect.

Description

Technical Field

[0001] The invention belongs to the technical field of architectural acoustics, and in particular relates to a method for acquiring an anti-reverberation speech signal suitable for a target architectural space. Background Art

[0002] In some special building spaces that require broadcasting (such as high-speed railway stations and airport terminals), the frequency of use in daily life has an increasing trend year by year. Whether it is daily notifications or emergency broadcasts, effective information communication is particularly critical in these spaces.

[0003] However, the unique shapes of these special architectural spaces and defects in acoustic design often lead to their unique acoustic characteristics, such as uneven sound field distribution and long reverberation time. The excessively long reverberation time can easily cause people in the space to not hear the voice signal clearly or even make people feel uncomfortable, which poses a great challenge to the effective transmission of information.

[0004] Moreover, indoor acoustic design is not only costly, but also faces great difficulties in making acoustic modifications in existing spaces.

[0005] Therefore, starting from the perspective of voice signals, this technology uses convolution technology to simulate the reverberation environment of special architectural spaces, and explores the optimization direction of voice signals suitable for special spatial reverberation conditions through human-computer collaborative learning to improve the clarity of voice signals. The goal of this method is to improve the clarity and efficiency of information transmission by adjusting and optimizing the emitted voice signals without expensive and complex acoustic equipment hardware modifications. In this way, better voice communication effects can be achieved at a very low cost in built architectural spaces. Summary of the invention

[0006] The purpose of the present invention is to overcome the deficiencies of the prior art and provide a method for acquiring an anti-reverberation speech signal suitable for a target building space.

[0007] The present invention is achieved through the following technical solutions:

[0008] A method for acquiring an anti-reverberation speech signal suitable for a target building space comprises the following steps:

[0009] Step 1: Collect the impulse response signal of the target building space;

[0010] Step 2: Build a real-time human-computer interaction simulation system;

[0011] The human-computer real-time interactive simulation system comprises: two stable sound environments, two sound playing devices, two sound collecting devices, a sound card and a computer; each stable sound environment is equipped with an experimenter, the sound playing device 1 and the sound collecting device 1 are arranged in the stable sound environment 1, the sound playing device 2 and the sound collecting device 2 are arranged in the stable sound environment 2, and the sound playing device 1, the sound collecting device 1, the sound playing device 2 and the sound collecting device 2 are all connected to the sound card; the sound card is also connected to the computer, and the computer is used to perform convolution processing on the original audio signal of the experimenter collected by the sound card and the impulse response signal of the target building space collected in step 1 to modulate them into a simulated audio signal to be played;

[0012] Step 3: Apply the human-computer real-time interaction simulation system to let two experimenters perform a dialogue task. During the dialogue task, experimenter A in the stable sound environment 1 makes a sound according to the dialogue task, and his sound is collected by the sound acquisition device 1 and the sound card as an original audio signal and sent to the computer. The computer performs convolution processing on the original audio signal emitted by experimenter A and the impulse response signal of the target building space collected in step 1, modulates it into an analog audio signal to be played, and sends it to the sound playback device 2 in the stable sound environment 2 through the sound card to play it to experimenter B. Experimenter B gives feedback based on his own auditory perception. If experimenter B cannot hear clearly, experimenter A corrects his own voice and makes the sound again. If experimenter B hears clearly, experimenter B makes a sound, and his sound is collected by the sound acquisition device 2 and the sound card. The collected audio is sent to a computer as an original audio signal. The computer performs convolution processing on the original audio signal emitted by the experimenter B and the impulse response signal of the target building space collected in step 1, modulates it into an analog audio signal to be played, and sends it to the sound playback device 1 in the stable sound environment 1 through the sound card to play it to the experimenter A. The experimenter A gives feedback according to his own auditory perception. If the experimenter A cannot hear clearly, the experimenter B corrects his own voice and re-produces the sound. If the experimenter A hears clearly, the experimenter A makes the sound again. This cycle allows the experimenters A and B to communicate effectively and continuously obtain feedback. According to the feedback, they continuously correct their own voices, so that the experimenters' voices tend to be stable, and both parties can hear clearly, thereby obtaining voice data in a stable voice state that the experimenters can hear clearly.

[0013] Step 4: By analyzing and learning the vocal characteristics of the speech data in the stable vocal state that the experimenters can hear clearly obtained in step 3, an anti-reverberation speech signal suitable for the target building space is obtained.

[0014] In the above technical solution, the pulse response signal of the target building space can be obtained through field measurement, using methods including but not limited to starting guns, electric sparks, balloons, electrical signals, etc.; it can also be obtained by software simulation, and the pulse response signal contains acoustic characteristic information of the target building space.

[0015] In the above technical scheme, the sound playing device 1 is connected to the first sound source output port of the sound card, and is used to send the audio signal to be played to the sound playing device 1 for playing; the sound collecting device 1 is connected to the first sound source input port of the sound card, and is used to collect the audio signal emitted by the experimenter A speaking in the stable sound environment 1 and transmit it to the sound card; the sound playing device 2 is connected to the second sound source output port of the sound card, and is used to send the audio signal to be played to the sound playing device 2 for playing; the sound collecting device 2 is connected to the second sound source input port of the sound card, and is used to collect the audio signal emitted by the experimenter B speaking in the stable sound environment 2 and transmit it to the sound card.

[0016] In the above technical solution, there are many ways to create the two stable sound environments, including but not limited to preparing two independent rooms, setting up sound barriers, placing soundproof rooms, etc.

[0017] In the above technical solution, the stable sound environment 1 and the stable sound environment 2 are not limited to a relatively independent relationship, but can also be a relationship of inclusion and being included. The computer and the sound card can also be placed in any sound environment or in other sound environments according to needs.

[0018] In the above technical scheme, the sound playing device, sound collecting device, computer, sound card and other devices are only functional requirements, without clear type regulations and model limitations. For example, the sound collecting device can use microphones, microphones, headsets and other devices with sound collecting functions; the connection method between the devices is not specifically limited, and it can be a wired connection or a wireless connection or other methods that ensure the transmission effect.

[0019] In the above technical solution, step 4 includes the following steps:

[0020] Step 4.1: First, extract features from the speech data obtained in step 3 in a stable state of speech that the experimenter can hear clearly;

[0021] Step 4.2: Then, using machine learning or deep learning methods, the extracted sound feature data is paired with the corresponding text to train an AI model that learns how to generate the corresponding human voice based on the given text;

[0022] Step 4.3: After the AI ​​model is trained, the given text is input into the model, and the model will generate corresponding dubbing voice signals. These generated dubbing voice signals have anti-reverberation performance suitable for the target building space, that is, anti-reverberation voice signals suitable for the target building space are obtained.

[0023] In the above technical solution, the anti-reverberation speech signal suitable for the target building space generated in step 4 is input into the broadcasting system of the target building space for playback, thereby achieving a good speech communication effect.

[0024] The advantages and beneficial effects of the present invention are:

[0025] Starting from the perspective of speech signals, the present invention uses convolution technology to simulate the reverberation environment of a special space, explores the optimization direction of speech signals suitable for reverberation conditions in a special space through human-computer collaborative learning, and improves the clarity of speech signals. The goal of this method is to improve the clarity and efficiency of information transmission by adjusting and optimizing the emitted speech signals without expensive and complex acoustic modifications. In this way, better speech transmission effects can be achieved in built spaces at a very low cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a structural diagram of a human-computer real-time interactive simulation system.

[0027] Figure 2 It is a schematic diagram of the convolution processing and modulation of the original audio signal and the impulse response signal of the target architectural space into the analog audio signal to be played.

[0028] Figure 3 It is another structural diagram of the human-computer real-time interactive simulation system.

[0029] Figure 4 It is another structural diagram of the human-computer real-time interactive simulation system.

[0030] Figure 5 It is another structural diagram of the human-computer real-time interactive simulation system.

[0031] Figure 6 It is another structural diagram of the human-computer real-time interactive simulation system.

[0032] Figure 7 It is another structural diagram of the human-computer real-time interactive simulation system.

[0033] For ordinary technicians in this field, other relevant drawings can be obtained based on the above drawings without any creative work. DETAILED DESCRIPTION

[0034] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention is further described below in conjunction with specific embodiments.

[0035] A method for acquiring an anti-reverberation speech signal suitable for a target building space comprises the following steps:

[0036] Step 1: First, simulate the acoustic environment of the target building space to restore the real broadcasting situation in the target building space and collect the impulse response signal of the target building space.

[0037] Specifically, it is necessary to collect the pulse response signal of the target building space. The pulse response signal can be obtained through field measurement, and the methods used include but are not limited to starting guns, electric sparks, balloons, electrical signals, etc.; it can also be obtained by software simulation. The pulse response signal contains acoustic characteristic information of the target building space.

[0038] Step 2: After obtaining the impulse response signal of the target building space and meeting the conditions for acoustic environment simulation, the next step is to build a real-time human-computer interaction simulation system.

[0039] like Figure 1 As shown, the human-computer real-time interactive simulation system includes: two stable sound environments, two sound playing devices, two sound collecting devices, a sound card and a computer. Each stable sound environment is equipped with an experimenter, the sound playing device 1 and the sound collecting device 1 are arranged in the stable sound environment 1, the sound playing device 2 and the sound collecting device 2 are arranged in the stable sound environment 2, and the sound playing device 1, the sound collecting device 1, the sound playing device 2 and the sound collecting device 2 are all connected to the sound card. Specifically, the sound playing device 1 is connected to the first sound source output port of the sound card, and is used to send the audio signal to be played to the sound playing device 1 for playing; the sound collecting device 1 is connected to the first sound source input port of the sound card, and is used to collect the audio signal emitted by the experimenter A speaking in the stable sound environment 1 and transmit it to the sound card; the sound playing device 2 is connected to the second sound source output port of the sound card, and is used to send the audio signal to be played to the sound playing device 2 for playing; the sound collecting device 2 is connected to the second sound source input port of the sound card, and is used to collect the audio signal emitted by the experimenter B speaking in the stable sound environment 2 and transmit it to the sound card. The sound card is also connected to a computer, which is used to perform convolution processing on the original audio signal of the experimenter collected by the sound card and the impulse response signal of the target building space collected in step 1 to modulate the original audio signal to be played (see attached Figure 2 ) to play it to another experimenter, so as to assign the acoustic characteristics of the target building space in the impulse response signal to the original audio signal of the sound, so as to simulate the experimenter making a sound in the reverberation condition of the target building space.

[0040] Furthermore, there are many ways to create two stable sound environments, including but not limited to preparing two separate rooms, setting up sound barriers, placing soundproof rooms, etc.

[0041] It should be noted that the two stable sound environments are intended to ensure that the sound in each space of the two sound environments is not affected by the other space and is not disturbed by noise. Figure 3 -Attached Figure 6 As shown, under the premise of ensuring that they do not interfere with each other, the stable sound environment 1 and the stable sound environment 2 are not limited to a relatively independent relationship, but can also be a relationship of inclusion and being included. The computer and the sound card can also be placed in any sound environment or in other sound environments according to needs.

[0042] It should also be noted that the sound playback device, sound collection device, computer, sound card and other devices are only functional requirements, and there is no clear type regulation and model limitation. For example, the sound collection device can use a microphone, a microphone, a headset and other devices with sound collection functions. There is no specific restriction on the connection method between the devices, which can be a wired connection, or a wireless connection or other methods that ensure the transmission effect.

[0043] Step 3: After the human-computer real-time interaction simulation system is built, the human-computer real-time interaction simulation system is used to let two experimenters perform a dialogue task. During the dialogue task, the two experimenters communicate and continuously obtain feedback, and continuously correct their own voices based on the feedback. Finally, the experimenters' voices tend to be stable and both parties can hear clearly, thereby obtaining voice data in a stable voice state that the experimenters can hear clearly.

[0044] For details, see the attached Figure 7, experimenter A in the stable sound environment 1 makes a sound according to the dialogue task, and his sound is collected by the sound collection device 1 and the sound card as an original audio signal and sent to the computer. The computer performs convolution processing on the original audio signal emitted by experimenter A and the impulse response signal of the target building space collected in step 1, modulates it into an analog audio signal to be played, and sends it to the sound playback device 2 in the stable sound environment 2 through the sound card to play it to experimenter B. Experimenter B gives feedback according to his own auditory perception. If experimenter B cannot hear clearly, experimenter A corrects his own voice and makes the sound again. If experimenter B hears clearly, experimenter B makes a sound, and his sound is collected by the sound collection device 2 and the sound card as an original audio signal and sent to the computer. The computer will The original audio signal emitted by the experimenter B is convolved with the impulse response signal of the target building space collected in step 1, and modulated into an analog audio signal to be played. The signal is sent to the sound playback device 1 in the stable sound environment 1 through the sound card to play to the experimenter A. The experimenter A gives feedback based on his own auditory perception. If the experimenter A cannot hear clearly, the experimenter B corrects his own voice and re-utters the sound. If the experimenter A hears clearly, the experimenter A utters the sound again. This cycle allows the experimenters A and B to communicate effectively and continuously obtain feedback. The experimenters continuously correct their own voices based on the feedback, so that the experimenters' voices tend to be stable and both parties can hear clearly, thereby obtaining voice data in a stable voice state that the experimenters can hear clearly.

[0045] It should be noted that the purpose of the dialogue task is to ensure that the two experimenters need to communicate effectively and continuously obtain feedback. The above-mentioned "hear clearly" and "did not hear clearly" are broad feedback purposes, not specific terms. Dialogue tasks come in various forms and there is no specific paradigm. In some embodiments, two experimenters perform a voice map task, that is, A says word 1 in sequence, B looks for the corresponding word 2 based on word 1 said by A, and feeds word 2 back to A, A looks for the corresponding word 3 based on word 2 fed back by B and gives feedback, and so on and so forth until the task is completed. The design method of the dialogue task can be flexibly adjusted according to specific needs.

[0046] Step 4: By analyzing and learning the vocal characteristics of the speech data in the stable vocal state that the experimenter can hear clearly obtained in step 3, an anti-reverberation speech signal suitable for the target building space is obtained. Specifically, step 4 includes the following steps:

[0047] Step 4.1: First, feature extraction is performed on the speech data in the stable vocalization state obtained in step 3 and which the experimenter can hear clearly, and sound features such as fundamental frequency, resonance peak, etc. are extracted.

[0048] Step 4.2: Then, using machine learning or deep learning methods, the extracted sound feature data is paired with the corresponding text to train an AI model that learns how to generate the corresponding human voice based on the given text.

[0049] Step 4.3: After the AI ​​model is trained, the given text can be input into the model, and the model will generate corresponding dubbing voice signals. These generated dubbing voice signals are very similar to the sound of the original audio data and have anti-reverberation performance suitable for the target building space, that is, anti-reverberation voice signals suitable for the target building space are obtained.

[0050] Step 5: The anti-reverberation speech signal suitable for the target building space generated in step 4 is input into the broadcasting system of the target building space for playback, thereby achieving a good speech communication effect.

[0051] The present invention is described above by way of example. It should be noted that, without departing from the core of the present invention, any simple deformation, modification or other equivalent replacement that can be made by those skilled in the art without inventive effort falls within the protection scope of the present invention.

Claims

1. A method for acquiring an anti-reverberation speech signal suitable for a target building space, characterized in that: The following steps are involved: Step 1: Collect the impulse response signal of the target building space; Step 2: Build a real-time human-computer interaction simulation system; The human-computer real-time interactive simulation system comprises: two stable sound environments, two sound playing devices, two sound collecting devices, a sound card and a computer; each stable sound environment is equipped with an experimenter, the sound playing device 1 and the sound collecting device 1 are arranged in the stable sound environment 1, the sound playing device 2 and the sound collecting device 2 are arranged in the stable sound environment 2, and the sound playing device 1, the sound collecting device 1, the sound playing device 2 and the sound collecting device 2 are all connected to the sound card; the sound card is also connected to the computer, and the computer is used to perform convolution processing on the original audio signal of the experimenter collected by the sound card and the impulse response signal of the target building space collected in step 1 to modulate them into a simulated audio signal to be played; Step 3: Apply the human-computer real-time interaction simulation system to let two experimenters perform a dialogue task. During the dialogue task, experimenter A in the stable sound environment 1 makes a sound according to the dialogue task, and his sound is collected by the sound acquisition device 1 and the sound card as an original audio signal and sent to the computer. The computer performs convolution processing on the original audio signal emitted by experimenter A and the impulse response signal of the target building space collected in step 1, modulates it into an analog audio signal to be played, and sends it to the sound playback device 2 in the stable sound environment 2 through the sound card to play it to experimenter B. Experimenter B gives feedback based on his own auditory perception. If experimenter B cannot hear clearly, experimenter A corrects his own voice and makes the sound again. If experimenter B hears clearly, experimenter B makes a sound, and his sound is collected by the sound acquisition device 2 and the sound card. The collected audio is sent to a computer as an original audio signal. The computer performs convolution processing on the original audio signal emitted by the experimenter B and the impulse response signal of the target building space collected in step 1, modulates it into an analog audio signal to be played, and sends it to the sound playback device 1 in the stable sound environment 1 through the sound card to play it to the experimenter A. The experimenter A gives feedback according to his own auditory perception. If the experimenter A cannot hear clearly, the experimenter B corrects his own voice and re-produces the sound. If the experimenter A hears clearly, the experimenter A makes the sound again. This cycle allows the experimenters A and B to communicate effectively and continuously obtain feedback. According to the feedback, they continuously correct their own voices, so that the experimenters' voices tend to be stable, and both parties can hear clearly, thereby obtaining voice data in a stable voice state that the experimenters can hear clearly. Step 4: By analyzing and learning the vocal characteristics of the speech data in the stable vocal state that the experimenter can hear clearly obtained in step 3, an anti-reverberation speech signal suitable for the target building space is obtained; Step 4 includes the following steps: Step 4.1: First, extract features from the speech data obtained in step 3 in a stable state of speech that the experimenter can hear clearly; Step 4.2: Then, using machine learning or deep learning methods, the extracted sound feature data is paired with the corresponding text to train an AI model that learns how to generate the corresponding human voice based on the given text; Step 4.3: After the AI ​​model is trained, the given text is input into the model, and the model will generate corresponding dubbing voice signals. These generated dubbing voice signals have anti-reverberation performance suitable for the target building space, that is, anti-reverberation voice signals suitable for the target building space are obtained.

2. The method for acquiring an anti-reverberation speech signal suitable for a target building space according to claim 1, characterized in that: The impulse response signal of the target building space is obtained through field measurement.

3. The method for acquiring an anti-reverberation speech signal suitable for a target building space according to claim 1, characterized in that: The impulse response signal of the target building space is obtained by software simulation.

4. The method for acquiring an anti-reverberation speech signal suitable for a target building space according to claim 1, characterized in that: The sound playing device 1 is connected to the first sound source output port of the sound card, and is used to send the audio signal to be played to the sound playing device 1 for playback; the sound collecting device 1 is connected to the first sound source input port of the sound card, and is used to collect the audio signal emitted by the experimenter A speaking in the stable sound environment 1 and transmit it to the sound card; the sound playing device 2 is connected to the second sound source output port of the sound card, and is used to send the audio signal to be played to the sound playing device 2 for playback; the sound collecting device 2 is connected to the second sound source input port of the sound card, and is used to collect the audio signal emitted by the experimenter B speaking in the stable sound environment 2 and transmit it to the sound card.

5. The method for acquiring an anti-reverberation speech signal suitable for a target building space according to claim 1, characterized in that: There are many ways to create the two stable sound environments, including: preparing two independent rooms, setting up sound barriers, and placing soundproof rooms.

6. The method for acquiring an anti-reverberation speech signal suitable for a target building space according to claim 1, characterized in that: The sound collection device adopts a microphone or a speaker.

7. The method for acquiring an anti-reverberation speech signal suitable for a target building space according to claim 1, characterized in that: The anti-reverberation speech signal suitable for the target building space generated in step 4 is input into the broadcasting system of the target building space for playback.

Images