Building management robot voice automatic identification method and system

Abstract

The invention relates to the technical field of management service robots, in particular to an automatic voice recognition method and system for a building management robot, and the method comprises the steps: analyzing the recognition difficulty of audio information according to the matching condition of the obtained audio information and the audio features of instruction audio information in audio data; when the matching degree of the collected audio information is high, instruction confirmation is sent to a manager, if the audio recognition is correct, a corresponding instruction signal is executed, if the audio recognition is not correct, the situation that the pronunciation of a sounding person is not standard is considered, the audio information is further recognized according to big data, and if the instruction confirmation feedback of the manager is not received, the corresponding instruction signal is executed; considering the condition that the manager does not send out an instruction and the collected audio information is noise, the building environment condition is analyzed through the matching degree of the audio information and then fed back through the manager, so that the condition of mistaken execution caused by insufficient recognition precision is avoided, resource waste is avoided, and the recognition accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of management service robots, and in particular to an automatic speech recognition method and system for a building management robot. Background Art

[0002] Building management robots are an important part of smart building systems. They integrate advanced technologies such as the Internet of Things, cloud computing, big data, and artificial intelligence to achieve intelligent, digital, and efficient building management. Robots can monitor and control lighting, air conditioning, elevators, security, and other equipment in buildings to achieve remote control and automated management. Existing technologies analyze the spectrum of voice information and compare and match the extracted voice features with key words. There is no need for complex voice model training and the recognition speed is fast. However, existing technologies do not take into account that in emergency evacuation situations, there is a lot of noise interference in the building, the flow of people is large, and the building echo will affect the robot's voice recognition, thereby affecting the execution efficiency of the building robot.

[0003] Chinese Patent Publication No.: CN112687269A discloses a method and system for automatic speech recognition of a building management robot, including: receiving voice information input from the outside; performing spectrum analysis on the received voice information; after analysis, comparing and matching the extracted voice features with the key words in the key word list, finding the key words with the smallest distance difference from the key word list as the recognition result and outputting it. The present invention adopts an intelligent voice system based on automatic speech recognition, and does not need to perform complex speech model training. It only needs to download the written recognition process program into the chip, and the recognition speed is extremely fast, and speech recognition of non-specific people (SI) can be realized. The module does not occupy embedded system resources, and is extremely convenient for system integration.

[0004] However, the prior art still has the following problems:

[0005] It was not taken into consideration that in the case of emergency evacuation, there would be a lot of noise interference in the building, a large flow of people and building echoes, which would affect the robot's voice recognition, thereby affecting the execution efficiency of the building robot. Summary of the invention

[0006] To this end, the present invention provides a method and system for automatic speech recognition of a building management robot, so as to overcome the problem in the prior art that in an emergency evacuation situation, there is a lot of noise interference in the building, a large flow of people and building echoes which will affect the robot's speech recognition, thereby affecting the execution efficiency of the building robot.

[0007] To achieve the above-mentioned purpose, the present invention provides a method for automatic speech recognition of a building management robot. The method comprises:

[0008] Step S1, constructing a command audio database, the command audio database including: command audio information and command signals corresponding to each command audio information;

[0009] Step S2, obtaining audio information, and analyzing the noise interference type of the audio information based on the matching degree between the audio information and the instruction audio information in the instruction audio database;

[0010] Step S3, determining the processing for the audio information based on the noise interference type, includes:

[0011] Obtain output feedback of the target audio information, analyze whether the speech recognition is qualified based on the feedback result, and re-recognize the audio or mark the audio;

[0012] Or, analyze the reasons for the mismatch of audio information, which may include: noise, echo, or insufficient audio acquisition sensitivity;

[0013] Step S4, calculating the matching degree between the processed audio information and the instruction audio information in the instruction audio database, and reanalyzing whether the audio information is the target audio information.

[0014] Further, determining the matching degree between the audio information and the instruction audio information in the instruction audio database includes:

[0015] Extract the audio features of the audio information and each command audio information respectively,

[0016] Count the number of audio features that are the same as the audio information of each command.

[0017] The instruction audio information having the largest number of audio features in common with the audio information is determined as the matching audio information,

[0018] Determine the number of audio features in the matching audio information that are identical to the audio information,

[0019] The ratio of the number of identical audio features to the total number of audio features in the audio information is calculated to obtain the matching degree.

[0020] Furthermore, the noise interference type of the audio information is analyzed based on the matching degree between the audio information and the instruction audio information in the instruction audio database, including: a weak noise interference type and a strong noise interference type.

[0021] Further, the determining of processing for the audio information based on the noise interference type includes:

[0022] Obtain output feedback of the target audio information, analyze whether the speech recognition is qualified based on the feedback result, re-recognize the audio or mark the audio, or analyze the reason why the audio information does not match.

[0023] Further, the analyzing whether the speech recognition is qualified based on the feedback result includes:

[0024] Determine that the speech recognition is qualified;

[0025] Determine if there is an accent problem, retrieve the corresponding accent type based on big data, and determine the content of the audio information based on the accent type, or,

[0026] Determine false triggering and mark the acquired audio information as noise.

[0027] Furthermore, the analyzing of the reasons for the mismatch of the audio information includes:

[0028] Draw a loudness-time curve of the audio information,

[0029] The audio information segment is determined based on the time interval between the maximum values ​​of the curve.

[0030] Determine that the two maximum values ​​are different audio information segments, or determine that the two maximum values ​​are the same audio information segment;

[0031] If the time interval between two adjacent maximum values ​​is less than the preset time interval standard threshold, the two maximum values ​​are determined to be the same audio information segment;

[0032] The number of audio information segments is counted, and the reasons for the mismatch of the audio information are analyzed based on the number of audio information segments.

[0033] Further, the analyzing the reasons for the mismatch of the audio information based on the number of audio information segments includes:

[0034] Determining whether there is noise in the audio information;

[0035] It is determined that there is an echo in the audio information, or that the audio acquisition sensitivity is insufficient.

[0036] Further, under the condition that it is determined that there is noise in the audio information, it is determined to re-collect the audio information, and the distance between the audio collection position and the sound source position is reduced based on the number of the audio information segments, wherein:

[0037] The reduction in the distance between the audio collection position and the sound source position is positively correlated with the number of audio information segments.

[0038] Further, under the condition that the audio information is determined to contain an echo, audio features in each audio information segment are extracted, and the type of the audio information segment is determined based on the audio features, including: a human voice information segment, an echo information segment, and a mixed information segment.

[0039] The audio features in the human voice information segment are extracted, and the audio features are matched with the instruction audio information in the instruction audio database.

[0040] The present invention provides a building management robot voice automatic recognition system. It includes:

[0041] A database construction module, which is used to construct a command audio database;

[0042] An audio acquisition module is used to acquire external audio information;

[0043] A matching module, which is connected to the database construction module and the audio acquisition module, and is used to calculate the matching degree between the audio information and the instruction audio information in the instruction audio database;

[0044] A noise interference analysis module, which is connected to the matching module and is used to determine the noise interference type of the audio information based on the matching degree calculated by the matching module;

[0045] A processing module is connected to the database construction module, the audio acquisition module, the matching module and the noise interference analysis module, and is used to determine the processing method for the audio information based on the noise interference type of the audio information, obtain the output feedback of the target audio information, analyze whether the speech recognition is qualified based on the feedback result, re-recognize the audio or mark the audio, and analyze the reason why the audio information does not match.

[0046] Compared with the prior art, the beneficial effect of the present invention lies in that the present invention analyzes the difficulty of recognizing audio information according to the matching situation of the audio characteristics of the acquired audio information and the audio of the command audio information in the audio data, and issues a command confirmation to the management personnel when the matching degree of the collected audio information is high. If the audio recognition is correct, the corresponding command signal is executed. If the audio recognition is incorrect, the audio information is further recognized according to the differences in accents in different places in the big data, considering that the speaker has non-standard pronunciation (accent, speaking too fast or too slow), if the command confirmation feedback from the management personnel is not received, considering that the management personnel has not issued any command and the collected audio information is noise, the building environment condition is first analyzed by the matching degree of the audio information, and then through the feedback from the management personnel, the situation of erroneous execution due to insufficient recognition accuracy is avoided, resource waste is avoided, and recognition accuracy is improved.

[0047] Furthermore, the present invention takes into account that when the command sound is issued, if the noise is after the command sound or the time when it intersects with the command sound is short, the influence on the command sound is small. When the noise and the command sound are in the same time period, the loudness of the two sounds is different. Considering that the intervals between the maximum values ​​of the loudness values ​​on the image will change due to different contents, when the two sounds intersect, the intervals between the maximum values ​​of the loudness are small. The intersection of the noise and the command sound is analyzed according to the time intervals between the maximum values, and divided into several audio segments (including audio segments with only command sounds, audio segments with only noise, and audio segments with both sounds). If various types of If there are a large number of audio segments, the impact of noise on the command sound will be greater. By shortening the distance between the sound collection end and the person issuing the command, the collected audio information can be easier to identify. If there are a small number of audio segments of each type, it is determined that there is an echo. Since the echo occurs after the command sound is issued, the audio information is classified and the audio features in the audio segment containing only the command sound are extracted and matched with the command audio information in the command audio database to reduce the secondary collection of audio information and improve execution efficiency. If there are very few audio information segments, it is considered that the sensitivity of the collection device is insufficient, so the sensitivity of the device is improved to improve the audio recognition efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a determination flow chart of the automatic speech recognition method of the building management robot of the present invention;

[0049] Figure 2 It is a structural block diagram of the automatic speech recognition system of the building management robot of the present invention;

[0050] Figure 3 A flow chart for determining the type of noise interference for analyzing audio information;

[0051] Figure 4 A decision flow chart for determining the processing of audio information. DETAILED DESCRIPTION

[0052] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0053] It should be pointed out that the data in this embodiment are obtained by comprehensive analysis and evaluation of the historical data of the system of the present invention in the six months before this determination and the corresponding historical determination results. It can be understood by those skilled in the art that the determination method of the system of the present invention for a single parameter mentioned above can be to select the value with the highest proportion as the preset standard parameter according to the data distribution, use weighted summation to use the obtained value as the preset standard parameter, substitute each historical data into a specific formula and use the value obtained by the formula as the preset standard parameter or other selection methods, as long as the system of the present invention can clearly define different specific situations in the single determination process through the obtained values.

[0054] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0055] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0056] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0057] See also Figure 1 As shown, it is a determination flow chart of the automatic speech recognition method of the building management robot of the present invention.

[0058] An embodiment of the present invention provides a method for automatic speech recognition of a building management robot, comprising:

[0059] Step S1, constructing a command audio database, the command audio database including: command audio information and command signals corresponding to each command audio information;

[0060] Step S2, obtaining audio information, and analyzing the noise interference type of the audio information based on the matching degree between the audio information and the instruction audio information in the instruction audio database;

[0061] Step S3, determining the processing for the audio information based on the noise interference type, includes:

[0062] Obtain output feedback of the target audio information, analyze whether the speech recognition is qualified based on the feedback result, and re-recognize the audio or mark the audio;

[0063] Or, analyze the reasons for the mismatch of audio information, which may include: noise, echo, or insufficient audio acquisition sensitivity;

[0064] Step S4, calculating the matching degree between the processed audio information and the instruction audio information in the instruction audio database, and reanalyzing whether the audio information is the target audio information.

[0065] See also Figure 2 As shown, it is a structural block diagram of the automatic speech recognition system of the building management robot of the present invention.

[0066] An embodiment of the present invention provides a building management robot voice automatic recognition system, comprising:

[0067] A database construction module, which is used to construct a command audio database;

[0068] An audio acquisition module is used to acquire external audio information;

[0069] A matching module, which is connected to the database construction module and the audio acquisition module, and is used to calculate the matching degree between the audio information and the instruction audio information in the instruction audio database;

[0070] A noise interference analysis module, which is connected to the matching module and is used to determine the noise interference type of the audio information based on the matching degree calculated by the matching module;

[0071] A processing module is connected to the database construction module, the audio acquisition module, the matching module and the noise interference analysis module, and is used to determine the processing method for the audio information based on the noise interference type of the audio information, obtain the output feedback of the target audio information, analyze whether the speech recognition is qualified based on the feedback result, re-recognize the audio or mark the audio, and analyze the reason why the audio information does not match.

[0072] Specifically, in this embodiment, the specific structures of the database construction module, matching module, noise interference analysis module and processing module are not limited, and they can be composed of logical components, and the logical components include a field editable processor, a computer and a microprocessor in the computer.

[0073] Specifically, in this embodiment, the specific structure of the audio acquisition module is not limited, and it can be a microphone or an acoustic sensor.

[0074] Specifically, determining the matching degree between the audio information and the instruction audio information in the instruction audio database includes:

[0075] Extract the audio features of the audio information and each command audio information respectively,

[0076] Count the number of audio features that are the same as the audio information of each command.

[0077] The instruction audio information having the largest number of audio features in common with the audio information is determined as the matching audio information,

[0078] Determine the number of audio features in the matching audio information that are identical to the audio information,

[0079] The ratio of the number of identical audio features to the total number of audio features in the audio information is calculated to obtain the matching degree.

[0080] See also Figure 3 As shown, it is a flow chart for determining the noise interference type for analyzing audio information.

[0081] Specifically, the analyzing the noise interference type of the audio information based on the matching degree between the audio information and the instruction audio information in the instruction audio database includes:

[0082] If the matching degree is greater than or equal to a preset matching degree standard threshold, the audio information is determined to be a weak noise interference type;

[0083] If the matching degree is less than the preset matching degree standard threshold, the audio information is determined to be of a strong noise interference type.

[0084] Specifically, in this embodiment, the preset matching standard threshold is obtained by pre-measurement, a number of noise-free audio information are collected, the matching degree of each audio information is calculated, and the matching degree mean is solved. The preset matching standard threshold is 0.8 to 0.95 times the matching degree mean.

[0085] See also Figure 4 As shown, it is a decision flow chart for determining audio information processing.

[0086] Specifically, determining the processing of the audio information based on the noise interference type includes:

[0087] If the audio information is of a weak noise interference type, obtaining output feedback of the target audio information, analyzing whether the speech recognition is qualified based on the feedback result, and re-recognizing the audio or marking the audio;

[0088] If the audio information is of a type with strong noise interference, the reason why the audio information does not match is analyzed.

[0089] Specifically, the analysis of whether the speech recognition is qualified based on the feedback result includes:

[0090] If the feedback result is confirmation, the speech recognition is judged to be qualified;

[0091] If the feedback result is wrong, it is determined that there is an accent problem, and the corresponding accent type is retrieved based on big data, and the content of the audio information is determined according to the accent type;

[0092] If there is no feedback result, it is determined to be a false trigger and the acquired audio information is marked as noise.

[0093] Specifically, the reasons for analyzing the mismatch of audio information include:

[0094] Draw a loudness-time curve of the audio information,

[0095] The audio information segment is determined based on the time interval between the maximum values ​​of the curve.

[0096] If the time interval between two adjacent maximum values ​​is greater than or equal to a preset time interval standard threshold, the two maximum values ​​are determined to be different audio information segments;

[0097] If the time interval between two adjacent maximum values ​​is less than the preset time interval standard threshold, the two maximum values ​​are determined to be the same audio information segment;

[0098] The number of audio information segments is counted, and the reasons for the mismatch of the audio information are analyzed based on the number of audio information segments.

[0099] Specifically, in this embodiment, the preset time interval standard threshold is obtained by pre-measurement, and audio information in a number of noise-free conditions is obtained, and a loudness-time image of each audio information is plotted. The time intervals between the maximum points in the image are counted, and the mean time interval is solved. The preset time interval standard threshold is 0.75 to 0.8 times the mean time interval.

[0100] Specifically, the analyzing the reasons for the mismatch of the audio information based on the number of audio information segments includes:

[0101] If the number of the audio information segments is greater than or equal to a first preset audio information segment number standard threshold, it is determined that noise exists in the audio information;

[0102] If the number of the audio information segments is less than the first preset audio information segment number standard threshold and is greater than or equal to the second preset audio information segment number standard threshold, determining that there is an echo in the audio information;

[0103] If the number of the audio information segments is less than the second preset audio information segment number standard threshold, it is determined that the audio collection sensitivity is insufficient.

[0104] Specifically, in this embodiment, the first preset standard threshold for the number of audio information segments and the second preset standard threshold for the number of audio information segments are obtained in advance, and a number of noisy audio information are obtained, the number of audio information segments in each audio information is counted, and the mean of the number of audio information segments is solved. The preset standard threshold for the number of audio information segments is 0.75 to 0.85 times the mean of the number of audio information segments, the first preset standard threshold for the number of audio information segments is 1.1 times the preset standard threshold for the number of audio information segments, and the second preset standard threshold for the number of audio information segments is 1.3 times the preset standard threshold for the number of audio information segments.

[0105] Specifically, under the condition that it is determined that noise exists in the audio information, it is determined to re-collect the audio information, and the distance between the audio collection position and the sound source position is reduced based on the number of the audio information segments, wherein:

[0106] The reduction in the distance between the audio collection position and the sound source position is positively correlated with the number of audio information segments.

[0107] In this embodiment, optionally,

[0108] The number of audio information segments is compared with a first preset audio information segment number standard threshold and a second preset audio information segment number standard threshold,

[0109] If the number of audio information segments is less than or equal to the first preset audio information segment number standard threshold, reducing the first distance to 0.15 to 0.2 times of the initial distance;

[0110] If the number of audio information segments is greater than the first preset audio information segment number standard threshold and less than or equal to the second preset audio information segment number standard threshold, then reducing the second distance, the second distance being 0.22 to 0.25 times the initial distance;

[0111] If the number of audio information segments is greater than the second preset audio information segment number standard threshold, the third distance is reduced, and the third distance is 0.3 to 0.35 times of the initial distance.

[0112] Specifically, under the condition that it is determined that there is an echo in the audio information, the audio features in each audio information segment are extracted, and the type of the audio information segment is determined based on the audio features, including: a human voice information segment, an echo information segment, and a mixed information segment.

[0113] The audio features in the human voice information segment are extracted, and the audio features are matched with the instruction audio information in the instruction audio database.

[0114] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0115] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for automatic speech recognition of a building management robot, characterized in that: include: Constructing a command audio database, the command audio database including: command audio information and command signals corresponding to each command audio information; Acquire audio information, and analyze the noise interference type of the audio information based on the matching degree between the audio information and the command audio information in the command audio database; The processing of the audio information is determined based on the type of noise interference, including: Obtain output feedback of the target audio information, analyze whether the speech recognition is qualified based on the feedback result, and re-recognize the audio or mark the audio; Or, analyze the reasons for the mismatch of audio information, which may include: noise, echo, or insufficient audio acquisition sensitivity; The matching degree between the processed audio information and the instruction audio information in the instruction audio database is calculated, and the audio information is re-analyzed to determine whether it is the target audio information.

2. The automatic speech recognition method for a building management robot according to claim 1, characterized in that: Determining a degree of matching between the audio information and instruction audio information in an instruction audio database includes: Extract the audio features of the audio information and each command audio information respectively, Count the number of audio features that are the same as the audio information of each command. The instruction audio information having the largest number of audio features in common with the audio information is determined as the matching audio information, Determine the number of audio features in the matching audio information that are identical to the audio information, The ratio of the number of identical audio features to the total number of audio features in the audio information is calculated to obtain the matching degree.

3. The automatic speech recognition method for a building management robot according to claim 1, characterized in that: The noise interference type of the audio information is analyzed based on the matching degree between the audio information and the instruction audio information in the instruction audio database, including: a weak noise interference type and a strong noise interference type.

4. The automatic speech recognition method for a building management robot according to claim 1, characterized in that: The step of determining processing for audio information based on the noise interference type includes: Obtain output feedback of the target audio information, analyze whether the speech recognition is qualified based on the feedback result, re-recognize the audio or mark the audio, or analyze the reason why the audio information does not match.

5. The automatic speech recognition method for a building management robot according to claim 1, characterized in that: The analyzing whether the speech recognition is qualified based on the feedback result includes: Determine that the speech recognition is qualified; Determine if there is an accent problem, retrieve the corresponding accent type based on big data, and determine the content of the audio information based on the accent type, or, Determine false triggering and mark the acquired audio information as noise.

6. The automatic speech recognition method for a building management robot according to claim 1, characterized in that: The reasons for the mismatch of the analyzed audio information include: Draw a loudness-time curve of the audio information, The audio information segment is determined based on the time interval between the maximum values ​​of the curve. Determine that the two maximum values ​​are different audio information segments, or determine that the two maximum values ​​are the same audio information segment; The number of audio information segments is counted, and the reasons for the mismatch of the audio information are analyzed based on the number of audio information segments.

7. The automatic speech recognition method for a building management robot according to claim 6, characterized in that: The analyzing the reasons for the mismatch of the audio information based on the number of audio information segments includes: Determining whether there is noise in the audio information; It is determined that there is an echo in the audio information, or that the audio acquisition sensitivity is insufficient.

8. The automatic speech recognition method for a building management robot according to claim 7, characterized in that: Under the condition that it is determined that there is noise in the audio information, it is determined to re-collect the audio information, and the distance between the audio collection position and the sound source position is reduced based on the number of the audio information segments, wherein: The reduction in the distance between the audio collection position and the sound source position is positively correlated with the number of audio information segments.

9. The automatic speech recognition method for a building management robot according to claim 7, characterized in that: Under the condition that the audio information is determined to contain an echo, audio features in each audio information segment are extracted, and the type of the audio information segment is determined based on the audio features, including: a human voice information segment, an echo information segment, and a mixed information segment, The audio features in the human voice information segment are extracted, and the audio features are matched with the instruction audio information in the instruction audio database.

10. A building management robot automatic speech recognition system implementing the method according to any one of claims 1 to 9, characterized in that: include: A database construction module, which is used to construct a command audio database; An audio acquisition module is used to acquire external audio information; A matching module, which is connected to the database construction module and the audio acquisition module, and is used to calculate the matching degree between the audio information and the instruction audio information in the instruction audio database; A noise interference analysis module, which is connected to the matching module and is used to determine the noise interference type of the audio information based on the matching degree calculated by the matching module; A processing module is connected to the database construction module, the audio acquisition module, the matching module and the noise interference analysis module, and is used to determine the processing method for the audio information based on the noise interference type of the audio information, obtain the output feedback of the target audio information, analyze whether the speech recognition is qualified based on the feedback result, re-recognize the audio or mark the audio, and analyze the reason why the audio information does not match.

Images