A method for deploying and accepting audio solutions

Abstract

The present invention discloses a method for deploying and accepting an audio solution, which includes a priori method and an acceptance method. The priori method of the present invention can determine whether the reverberation time meets the needs of the product according to the reverberation time obtained by the priori method. If it does not meet the needs of the product, the environment can be improved according to a subsequent formula to shorten the reverberation time. The acceptance method of the present invention allows engineers to determine which link has a problem according to the above scores, and determine which link can be improved and optimized according to their own knowledge accumulation and engineering, and finally the audio solution can be well deployed and accepted according to this method.

Description

Technical Field

[0001] The present invention relates to the field of audio engineering, and in particular to a method for deploying and accepting an audio solution. Background Art

[0002] Many engineering projects are currently inseparable from audio and video technology. Among them, audio is greatly affected by the scene environment. Many project deployments do not have scientific deployment rules. Most projects are deployed based on their own knowledge accumulation and project experience, and acceptance is judged based on their own subjective assumptions. This makes it difficult to meet the customer's acceptance standards and explain the sound quality to the customer. At the same time, it is difficult to locate the problem point when solving the problem.

[0003] The prior art method for calculating reverberation time requires the use of a theoretical formula: Calculate, where V is the building volume, S is the indoor surface area, The sound absorption coefficient is the surface area. In addition, the sound absorption coefficient is extremely complex and is related to the material of the scene. It is often missed or miscalculated. At present, the acceptance is mainly completed through the subjective feelings of engineers. It is impossible to explain the voice quality with reason and evidence, and it is difficult to locate the problem. Summary of the invention

[0004] The object of the present invention is to provide a method for deployment and acceptance of an audio solution.

[0005] The technical solution adopted by the present invention is:

[0006] A method for deploying and accepting an audio solution includes a priori testing part and an acceptance part, and the specific steps are as follows:

[0007] Prior part:

[0008] Step 11: Play white noise in the scene where audio deployment is required;

[0009] Step 12, collecting audio data by testing the sound pickup device;

[0010] Step 13, calculating the reverberation time of the audio data by a computing device;

[0011] Step 14, determine whether the reverberation time is not greater than the tolerable maximum reverberation time; if so, determine that the current scene meets the requirements and output the reverberation time; otherwise, determine that the scene requirements are not met and provide scene rectification suggestions;

[0012] Furthermore, in step 11, a portable amplifier is used to play white noise, such as a Bluetooth speaker.

[0013] Furthermore, the sound pickup device is a microphone, and the computing device is a laptop computer.

[0014] Furthermore, the reverberation time in step 13 is: the time required for the density of average sound energy to decay from the original value to one millionth of the original value after the sound source stops emitting sound after the sound in the room reaches a stable state.

[0015] Furthermore, in step 13, the time required for the sound level in the room to decay by 60 db after the sound source occurs is used as the reverberation time.

[0016] Furthermore, in step 14, when the reverberation time does not meet the scene requirements, the sound-absorbing material of the corresponding area is increased according to the size of the scene. The specific area calculation formula is as follows:

[0017]

[0018] Where V is the building volume, S 60 The indoor surface area, S is the area that needs to be increased; Is the sound absorption coefficient, different materials have different absorption coefficients.

[0019] Acceptance part: The acceptance plan means that after the entire audio solution is built, we need to accept the entire audio solution, mainly whether there is wiring, whether the installation position of the microphone is reasonable, or other unexpected problems (such as interference with other audio solutions); at the same time, it can also give customers a satisfactory score. When encountering problems, the problems can be judged based on the output; in normal projects, the acceptance of audio solutions mainly includes two forms: one is single-ended, and the other is the acceptance of both ends. Single-ended and dual-ended mainly involve a network data transmission in the middle. The specific acceptance steps are as follows:

[0020] Step 21, determining whether the sound source is a standard sound; if so, executing step 22; otherwise, downloading the standard sound and playing it through a playback device and executing step 22;

[0021] Step 22, assessing the current voice quality to determine the voice quality level;

[0022] Step 23, determine whether it is a single-end acceptance or a double-end acceptance; when it is a single-end acceptance, execute step 26; when it is a double-end acceptance, execute step 24;

[0023] Step 24, evaluate the current network quality to determine whether the network quality affects the current voice quality level; if the main network factor affecting the audio is packet loss, the packet loss rate of the network can be checked in a query manner to determine whether it affects the voice quality.

[0024] Step 25, determining whether the network quality test result meets the standard; if yes, executing step 26; otherwise, executing step 27;

[0025] Step 26, outputting the result or converting it into a set score;

[0026] Step 27, display the score and indicate the deficiencies.

[0027] Furthermore, in step 22, the current voice quality is assessed by a voice quality expert algorithm trained based on a deep neural network learning algorithm to determine a voice quality level.

[0028] The voice quality expert algorithm is formulated based on a deep neural network learning algorithm and is trained with a large amount of human voice data (select which training algorithm to collect data based on the scenario). The input is a normal voice signal, and its file must be a .wav file. If not, a conversion tool is required for conversion; the output is a MOS score, which is the average subjective opinion score and can be a decimal. The basic levels are: 1. 4-5 (inclusive) points are excellent (excelent), very good, clear hearing, small delay, and smooth communication; 2. 3-4 (not included) points are good (good), slightly worse, clear hearing, small delay, communication is not smooth, and there is some noise; 3. 2-3 (not included) points are fair (fair), not very clear, with a certain delay, and communication is possible; 4. 1-2 (not included) points are poor (poor), not very clear, with a large delay, and repeated communication; 5. Below 1 point is very bad (bad), extremely bad, incomprehensible, with a large delay, and poor communication.

[0029] The specific steps of the voice quality expert algorithm are:

[0030] Step 22-1, acquiring a speech signal and extracting a Mel spectrum;

[0031] Step 22-2, segmenting the Mel spectrum;

[0032] Step 22-3, extracting the segmented Mel spectrum by speech framing to obtain frame-by-frame features;

[0033] Framing is an existing technology. General speech processing requires the speech to be framed (that is, processing a segment of speech by a few seconds, because speech is short-term invariant to analyze the characteristics of speech).

[0034] Step 22-4, inputting the frame-by-frame features into a deep network model for evaluating speech quality and calculating the corresponding weights respectively;

[0035] Specifically, the deep neural network learning algorithm model can adopt existing speech quality evaluation technologies, such as MOSNet, NISQA, AutoMOS, QuanlityNet, etc.

[0036] Step 22-5: Perform pooling processing to obtain the MOS score.

[0037] Furthermore, in step 22-5, five MOS scores are finally obtained, namely: expert final score MOS, speech size MOS, speech correlation MOS, speech continuity MOS and speech noise MOS;

[0038] The final expert score MOS is the score that the experts can give to this solution.

[0039] The voice volume MOS is used to test whether the sound volume of the solution meets the requirements. The score is obtained by comparing the volume of the voice input with the comfort level of the average person. The larger the better, the smaller the better.

[0040] Speech relevance MOS is used to test whether the sound of the solution is relevant speech input. Generally, the standard is human voice. If it is not human voice, the value is relatively low, which will affect the final score.

[0041] Speech continuity MOS mainly tests whether the solution is continuous sound input; if there is a discontinuous situation, the value will be scored according to the degree of continuity;

[0042] Speech noise MOS refers to the amount of noise mixed in the speech input. The greater the noise, the smaller the value.

[0043] The present invention adopts the above technical solution. Before the audio equipment (microphone, speaker, audio processor and platform and other equipment) is installed, the reverberation time of the environment needs to be tested, and the matching degree between the environment and the equipment is judged by the reverberation time. At the same time, professional improvement suggestions can be given; after the audio equipment is deployed, normal Mandarin (or standard sound) is used to form the sound source for playback, and the microphone collects the sound to the algorithm platform for evaluation, and finally the output score is judged whether it meets the standard. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments;

[0045] Figure 1 It is a schematic diagram of the process framework of the a priori method of the present invention;

[0046] Figure 2 It is a flow chart of the reverberation time algorithm of the present invention;

[0047] Figure 3 This is a schematic diagram of the process when the present invention adopts single-end acceptance;

[0048] Figure 4 This is a schematic diagram of the process when the present invention adopts double-end acceptance;

[0049] Figure 5 It is a flow chart of the speech quality expert algorithm of the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0051] like Figures 1 to 5 As shown in one, the present invention discloses a method for deploying and accepting an audio solution, which includes a priori testing part and an acceptance part:

[0052] 1) A priori method: With the help of peripheral devices, including portable microphones, laptops, white noise, and Bluetooth speakers; the Bluetooth speaker plays white noise, which is collected by the microphone and the laptop calculates the reverberation time. Based on the reverberation time, it is determined whether the room meets the equipment matching requirements. If not, modification suggestions are given.

[0053] 2) Acceptance method: After the audio equipment is deployed, the deployment engineer speaks standard Mandarin (if not, do not play the previously downloaded standard sound through the Bluetooth speaker), the sound is played to the microphone on site and input into the notebook (or other computer carrier), and then converted into a .wav file through the voice format and input into the acceptance algorithm, and finally several intuitive scores are obtained, which is convenient for project acceptance judgment and problem location. The specific steps are as follows:

[0054] Prior part: such as Figure 1 Or as shown in 2:

[0055] Step 11: Play white noise in the scene where audio deployment is required;

[0056] Step 12, collecting audio data by testing the sound pickup device;

[0057] Step 13, calculating the reverberation time of the audio data by a computing device;

[0058] Step 14, determine whether the reverberation time is not greater than the tolerable maximum reverberation time; if so, determine that the current scene meets the requirements and output the reverberation time; otherwise, determine that the scene requirements are not met and provide scene rectification suggestions;

[0059] Furthermore, in step 11, a portable amplifier is used to play white noise, such as a Bluetooth speaker.

[0060] Furthermore, the sound pickup device is a microphone, and the computing device is a laptop computer.

[0061] Furthermore, the reverberation time in step 13 is: the time required for the density of average sound energy to decay from the original value to one millionth of the original value after the sound source stops emitting sound after the sound in the room reaches a stable state.

[0062] Furthermore, in step 13, the time required for the sound level in the room to decay by 60 db after the sound source occurs is used as the reverberation time.

[0063] Furthermore, in step 14, when the reverberation time does not meet the scene requirements, the sound-absorbing material of the corresponding area is increased according to the size of the scene. The specific area calculation formula is as follows:

[0064]

[0065] Where V is the building volume, S 60 The indoor surface area, S is the area that needs to be increased; Is the sound absorption coefficient, different materials have different absorption coefficients.

[0066] Acceptance part: The acceptance plan means that after the entire audio solution is built, we need to accept the entire audio solution, mainly to check whether the wiring, the installation position of the microphone is reasonable, or other unexpected problems (such as interference with other audio solutions); at the same time, we can also give customers a satisfactory score. When encountering problems, we can judge the problems based on the output; in normal projects, the acceptance of audio solutions mainly includes two forms: one is single-ended, and the other is the acceptance of both ends. Single-ended and dual-ended mainly involve a network data transmission in the middle. For example Figure 3 and 4 The specific acceptance steps are as follows:

[0067] Step 21, determining whether the sound source is a standard sound; if so, executing step 22; otherwise, downloading the standard sound and playing it through a playback device and executing step 22;

[0068] Step 22, assessing the current voice quality to determine the voice quality level;

[0069] Step 23, determine whether it is a single-end acceptance or a double-end acceptance; when it is a single-end acceptance, execute step 26; when it is a double-end acceptance, execute step 24;

[0070] Step 24, evaluate the current network quality to determine whether the network quality affects the current voice quality level; if the main network factor affecting the audio is packet loss, the packet loss rate of the network can be checked in a query manner to determine whether it affects the voice quality.

[0071] Step 25, determining whether the network quality test result meets the standard; if yes, executing step 26; otherwise, executing step 27;

[0072] Step 26, outputting the result or converting it into a set score;

[0073] Step 27, display the score and indicate the deficiencies.

[0074] Furthermore, in step 22, the current voice quality is assessed by a voice quality expert algorithm trained based on a deep neural network learning algorithm to determine a voice quality level.

[0075] The voice quality expert algorithm is developed based on a deep neural network learning algorithm and is trained with a large amount of human voice data (select which training algorithm to use based on the scenario). The input is a normal voice signal, which must be a .wav file. If not, a conversion tool is required for conversion. The output is a MOS score, which is the average subjective opinion score and can be a decimal. The basic levels are:

[0076] 1. 4-5 (inclusive) Excellent: Very good, can hear clearly, with little delay, and can communicate fluently;

[0077] 2. 3-4 (not included) Good (slightly worse), clear hearing, small delay, communication is not smooth, and there is some noise;

[0078] 3. 2-3 (not included) points: fair, not very clear, with some delay, but able to communicate;

[0079] 4. 1-2 (not included) score difference (poor) barely understandable, large delay, repeated communication

[0080] 5. Below 1 point is very bad (bad), extremely bad, incomprehensible, large delay, and poor communication.

[0081] like Figure 5 As shown, the specific steps of the voice quality expert algorithm are:

[0082] Step 22-1, acquiring a speech signal and extracting a Mel spectrum;

[0083] Step 22-2, segmenting the Mel spectrum;

[0084] Step 22-3, extracting the segmented Mel spectrum by speech framing to obtain frame-by-frame features;

[0085] Framing is an existing technology. General speech processing requires the speech to be framed (that is, processing a segment of speech by a few seconds, because speech is short-term invariant to analyze the characteristics of speech).

[0086] Step 22-4, inputting the frame-by-frame features into a deep network model for evaluating speech quality and calculating the corresponding weights respectively;

[0087] Specifically, the deep neural network learning algorithm model can adopt existing speech quality evaluation technologies, such as MOSNet, NISQA, AutoMOS, QuanlityNet, etc.

[0088] Step 22-5: Perform pooling processing to obtain the MOS score.

[0089] Furthermore, in step 22-5, five MOS scores are finally obtained, namely: expert final score MOS, speech size MOS, speech correlation MOS, speech continuity MOS and speech noise MOS;

[0090] The final expert score MOS is the score that the experts can give to this solution.

[0091] The voice volume MOS is used to test whether the sound volume of the solution meets the requirements. The score is obtained by comparing the volume of the voice input with the comfort level of the average person. The larger the better, the smaller the better.

[0092] Speech relevance MOS is used to test whether the sound of the solution is relevant speech input. Generally, the standard is human voice. If it is not human voice, the value is relatively low, which will affect the final score.

[0093] Speech continuity MOS mainly tests whether the solution is continuous sound input; if there is a discontinuous situation, the value will be scored according to the degree of continuity;

[0094] Speech noise MOS refers to the amount of noise mixed in the speech input. The greater the noise, the smaller the value.

[0095] The a priori method of the present invention can determine whether the reverberation time is in accordance with the needs of the product according to the a priori method. If it is not satisfied, the environment can be improved according to the subsequent formula to shorten the reverberation time to meet the needs of the product. The acceptance method of the present invention allows engineers to determine which link has a problem according to the above-mentioned scores, and determine which link can be improved and optimized according to their own knowledge accumulation and engineering. Finally, the audio solution can be well deployed and accepted according to this method.

[0096] The present invention adopts the above technical solution. Before the audio equipment (microphone, speaker, audio processor and platform and other equipment) is installed, the reverberation time of the environment needs to be tested, and the matching degree between the environment and the equipment is judged by the reverberation time. At the same time, professional improvement suggestions can be given; after the audio equipment is deployed, normal Mandarin (or standard sound) is used to form the sound source for playback, and the microphone collects the sound to the algorithm platform for evaluation, and finally the output score is judged whether it meets the standard.

[0097] Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. In the absence of conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of the present application is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians of the art without making creative work are within the scope of protection of the present application.

Claims

1. A method for deploying and accepting an audio solution, characterized in that: It includes a priori testing part and acceptance part. The specific steps are as follows: Prior part: Step 11: Play white noise in the scene where audio deployment is required; Step 12, collecting audio data by testing the sound pickup device; Step 13, calculating the reverberation time of the audio data by a computing device; Step 14, determine whether the reverberation time is not greater than the tolerable maximum reverberation time; if so, determine that the current scene meets the requirements and output the reverberation time; otherwise, determine that the scene requirements are not met and provide scene rectification suggestions; In step 14, when the reverberation time does not meet the scene requirements, the sound-absorbing material of the corresponding area is increased according to the size of the scene. The specific area calculation formula is as follows: ； Where V is the building volume, The indoor surface area, S is the area that needs to be increased; is the sound absorption coefficient, and different materials have different absorption coefficients; Acceptance part: Step 21, determining whether the sound source is a standard sound; if so, executing step 22; otherwise, downloading the standard sound and playing it through a playback device and executing step 22; Step 22, assessing the current voice quality to determine the voice quality level; Step 23, determine whether it is a single-end acceptance or a double-end acceptance; when it is a single-end acceptance, execute step 26; when it is a double-end acceptance, execute step 24; Step 24, evaluating the current network quality to determine whether the network quality affects the current voice quality level; Step 25, determining whether the network quality test result meets the standard; if yes, executing step 26; otherwise, executing step 27; Step 26, outputting the result or converting it into a set score; Step 27, display the score and indicate the deficiencies.

2. The method for deploying and accepting an audio solution according to claim 1, characterized in that: In step 11, a portable amplifier is used to play white noise.

3. The method for deploying and accepting an audio solution according to claim 1, characterized in that: The sound pickup device is a microphone, and the computing device is a laptop computer.

4. The method for deploying and accepting an audio solution according to claim 1, characterized in that: The reverberation time in step 13 is: the time required for the density of average sound energy to decay from the original value to one millionth of the original value after the sound source stops emitting sound after the sound in the room reaches a stable state.

5. The method for deploying and accepting an audio solution according to claim 4, characterized in that: In step 13, the time required for the sound level in the room to decay by 60 dB after the sound source is generated is taken as the reverberation time.

6. The method for deploying and accepting an audio solution according to claim 1, characterized in that: In step 22, the current voice quality is evaluated by a voice quality expert algorithm trained based on a deep neural network learning algorithm to determine the voice quality level; the voice quality expert algorithm is formulated based on a deep neural network learning algorithm and is trained through a large amount of human voice data to collect data according to the scenario to select which training algorithm to use; the input is a normal voice signal and the output is a MOS score, which is the average subjective opinion score.

7. The method for deploying and accepting an audio solution according to claim 6, characterized in that: The specific steps of the voice quality expert algorithm are: Step 22-1, acquiring a speech signal and extracting a Mel spectrum; Step 22-2, segmenting the Mel spectrum; Step 22-3, extracting the segmented Mel spectrum by speech framing to obtain frame-by-frame features; Step 22-4, inputting the frame-by-frame features into a deep network model for evaluating speech quality and calculating the corresponding weights respectively; Step 22-5: Perform pooling processing to obtain the MOS score.

8. The method for deploying and accepting an audio solution according to claim 7, characterized in that: In step 22-5, five MOS scores are finally obtained, namely: expert final score MOS, speech size MOS, speech correlation MOS, speech continuity MOS and speech noise MOS; The final expert score MOS is the score that the experts can give to this solution. The voice volume MOS is used to test whether the sound volume of the solution meets the requirements. The score is obtained by comparing the volume of the voice input with the comfort level of the average person. The larger the better, the smaller the better. Speech relevance MOS is used to test whether the sound of the solution is relevant speech input. Generally, the standard is human voice. If it is not human voice, the value is relatively low, which will affect the final score. Speech continuity MOS mainly tests whether the solution is continuous sound input; if there is a discontinuous situation, the value will be scored according to the degree of continuity; Speech noise MOS refers to the amount of noise mixed in the speech input. The greater the noise, the smaller the value.

Images