Air conditioner control method and system based on accent speech recognition

Abstract

An air conditioner control method and system based on accent speech recognition comprises the following steps that a language adjusting command is set, and a plurality of language databases are constructed according to the language adjusting command and various dialects; wherein the language adjustment command comprises a wake-up command and a control command; acquiring an audio of a wake-up command input by a user, and uploading the audio to a cloud; converting the audio into a first pinyin, finding out a target language database from a language database according to the first pinyin, and loading a language control configuration file associated with dialects in the target language database into the air conditioner. According to the method, the audio is directly converted into the first pinyin, and more details can be exposed on the pinyin level, so that matching can be performed more accurately during matching, and the recognition efficiency of dialects is improved. Therefore, the user can interact with the equipment in a more natural and familiar manner, so that the operation convenience and satisfaction of the user are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of air conditioning regulation, and in particular to an air conditioning control method and system based on accented speech recognition. Background Art

[0002] With the rapid development of smart technology, smart voice air conditioners, as an emerging force in the field of smart homes, are penetrating into people's daily lives at an unprecedented rate, greatly improving the convenience and comfort of home life.

[0003] However, despite its continued rise in popularity, a realistic challenge that cannot be ignored has gradually surfaced: the current voice air conditioning system mainly relies on the recognition ability of standard Mandarin, and this design has certain limitations in its wide applicability. For example, in some dialect areas or for the elderly and children who are learning languages, they may find it difficult to effectively control the air conditioner through voice commands due to differences in accents, different language habits, or lack of proficiency in Mandarin. This has greatly reduced the intelligent experience of voice air conditioning. Summary of the invention

[0004] In view of the above-mentioned defects, the purpose of the present invention is to propose an air-conditioning control method and system based on accented speech recognition to solve the problem in the prior art that it is difficult to use dialects to call the air conditioner.

[0005] To achieve this purpose, the present invention adopts the following technical solution: an air conditioning control method based on accented speech recognition, comprising the following steps:

[0006] Setting a language adjustment command, and constructing a number of language databases for the language adjustment command and various dialects;

[0007] The language adjustment command includes a wake-up command and a control command;

[0008] Acquire the audio of the wake-up command input by the user, and upload the audio to the cloud;

[0009] The audio is converted into a first pinyin, a target language database is found from a language database according to the first pinyin, a language control configuration file associated with a dialect in the target language database is generated, and the language control configuration file is loaded into the air conditioner.

[0010] Preferably, the step of converting the audio into the first pinyin is as follows:

[0011] Preprocessing the audio of the input wake-up command to obtain processed data;

[0012] The MFCC speech feature sequence is extracted from the processed data to obtain the speech feature sequence, and the speech feature sequence is input into the existing acoustic layer recognition module. The recognition result in the form of the first pinyin grid is generated through the search algorithm of the acoustic layer recognition module.

[0013] Preferably, the step of finding the target language database according to the first pinyin is as follows:

[0014] Filtering at least one first database from the language database according to the length of the first pinyin;

[0015] Filtering at least one second database from the first database according to the key field in the first pinyin;

[0016] The first pinyin is used to match the second pinyins of the wake-up commands in multiple second databases respectively, and the second database with the highest matching value is used as the target language database.

[0017] Preferably, the steps of using the first pinyin to match the second pinyin are as follows:

[0018] The first pinyin and the second pinyin are encoded by one-hot, and vectorized to obtain a first vector and a second vector respectively;

[0019] The similarity between the first vector and the second vector is calculated, and the second vector with the highest similarity is obtained, so that the corresponding second database is used as the target language database.

[0020] Preferably, the method further comprises the following steps:

[0021] Get the time when the language control configuration file is loaded as the first time;

[0022] Adding the threshold time to the first time to obtain a second time;

[0023] It is determined whether the language control configuration file is called within the second time. If it is called, the calling time is obtained and updated to the first time; if it is not called, the language control configuration file in the air conditioner is deleted.

[0024] An air conditioning control system based on accented speech recognition, using the air conditioning control method based on accented speech recognition, characterized in that it includes a setting module, an audio acquisition module and a cloud module;

[0025] The setting module is used to set language adjustment commands, and to construct a number of language databases for the language adjustment commands and various dialects;

[0026] The audio acquisition module is used to acquire the audio of the wake-up command input by the user and upload the audio to the cloud;

[0027] The cloud module is used to convert the audio into a first pinyin, find a target language database according to the first pinyin, associate a language control configuration file with a dialect in the target language database, and load the language control configuration file into the air conditioner.

[0028] Preferably, the cloud module includes a conversion submodule;

[0029] The conversion submodule is used to pre-process the audio of the input wake-up command to obtain processed data;

[0030] The MFCC speech feature sequence is extracted from the processed data to obtain the speech feature sequence, and the speech feature sequence is input into the existing acoustic layer recognition module. The recognition result in the form of the first pinyin grid is generated through the search algorithm of the acoustic layer recognition module.

[0031] Preferably, the cloud module includes a matching submodule;

[0032] The matching submodule is used to filter out at least one first database from the language database according to the length of the first pinyin;

[0033] Filtering at least two types of databases from the first database according to the key fields in the first pinyin;

[0034] The first pinyin is used to match the second pinyins of the language adjustment commands in multiple second databases respectively, and the second database with the highest matching value is used as the target language database.

[0035] Preferably, the matching submodule includes a calculation unit;

[0036] The calculation unit is used to encode the first pinyin and the second pinyin respectively through one-hot, and vectorize them to obtain a first vector and a second vector respectively;

[0037] The similarity between the first vector and the second vector is calculated, and the second vector with the highest similarity is obtained, so that the corresponding second database is used as the target language database.

[0038] Preferably, it also includes a clearing module;

[0039] The clearing module is used to obtain the time when the language control configuration file is loaded as the first time;

[0040] Adding the threshold time to the first time to obtain a second time;

[0041] It is determined whether the language control configuration file is called within the second time. If it is called, the calling time is obtained and updated to the first time; if it is not called, the language control configuration file in the air conditioner is deleted.

[0042] One of the above technical solutions has the following advantages or beneficial effects: the present invention directly converts the audio into the first pinyin, which can reveal more details at the pinyin level, so that the matching can be more accurate during matching, and the recognition efficiency of the dialect is improved. It enables users to interact with the device in a more natural and familiar way, thereby greatly improving the user's operational convenience and satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is a flow chart of an embodiment of the method of the present invention.

[0044] Figure 2 It is a schematic diagram of the structure of an embodiment of the system of the present invention. DETAILED DESCRIPTION

[0045] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions from beginning to end. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0046] In the description of the embodiments of the present invention, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0047] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, the meaning of "plurality" is two or more. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0048] like Figures 1-2 As shown, an air conditioning control method based on accented speech recognition comprises the following steps:

[0049] Setting a language adjustment command, and constructing a number of language databases for the language adjustment command and various dialects;

[0050] The language adjustment command includes a wake-up command and a control command;

[0051] Acquire the audio of the wake-up command input by the user, and upload the audio to the cloud;

[0052] The audio is converted into a first pinyin, a target language database is found from a language database according to the first pinyin, a language control configuration file associated with a dialect in the target language database is generated, and the language control configuration file is loaded into the air conditioner.

[0053] In order to enable the voice air conditioning system to be applied to different dialect areas, the present invention first sets a language adjustment command, wherein the language adjustment command includes a wake-up command and a control command; the wake-up command and the control command are set using the corresponding dialect in different dialects, for example, Mandarin, Minnan dialect, Shandong dialect, and Cantonese are used to set the corresponding wake-up command and the control command respectively. When in use, the user needs to use the wake-up command to call the corresponding air conditioner first; a specific wake-up word can be added to the wake-up command to avoid misjudgment, for example, the wake-up command can be "Xiaofan, please start the voice block mode".

[0054] After receiving the audio, the air conditioner will upload the audio to the cloud. Since the processor in the air conditioner is relatively backward and cannot perform more complex calculations, subsequent matching processing is required in the cloud. The audio is converted into the first pinyin through cloud processing. Since the comparison is made directly through the audio, it is difficult to use the audio directly for matching because the user speaks in a different way every time. In the prior art, the audio is converted into text and then matched through text. However, this method is difficult to match and identify in relatively similar dialects, such as Minnan dialect and Chaoshan dialect. Therefore, the present invention directly converts the audio into the first pinyin, which can reveal more details at the pinyin level, so that the matching can be more accurate during matching, thereby improving the recognition efficiency of dialects. It enables users to interact with the device in a more natural and familiar way, thereby greatly improving the user's operational convenience and satisfaction.

[0055] Preferably, the step of converting the audio into the first pinyin is as follows:

[0056] Preprocessing the audio of the input wake-up command to obtain processed data;

[0057] The MFCC speech feature sequence is extracted from the processed data to obtain the speech feature sequence, and the speech feature sequence is input into the existing acoustic layer recognition module. The recognition result in the form of the first pinyin grid is generated through the search algorithm of the acoustic layer recognition module.

[0058] The pre-processing is to frame the audio signal using a sliding window, and there is usually a certain overlap between the windows, and then the overlapping parts are deleted to ensure the continuity between the frames. In the subsequent pinyin conversion, repeated syllables can be avoided to affect the matching results.

[0059] In addition, in the present invention, MFCC is used to extract speech feature sequences, and MFCC can better simulate the auditory characteristics of the human ear and efficiently and accurately extract features from audio signals, so that it still has good recognition performance in a noisy environment and enhances the robustness of the system.

[0060] Preferably, the step of finding the target language database according to the first pinyin is as follows:

[0061] Filtering at least one first database from the language database according to the length of the first pinyin;

[0062] Filtering at least one second database from the first database according to the key field in the first pinyin;

[0063] The first pinyin is used to match the second pinyins of the wake-up commands in multiple second databases respectively, and the second database with the highest matching value is used as the target language database.

[0064] When the user speaks the wake-up command in a dialect, erhua sounds or other modal particles may appear in the dialect. At this time, the length of the first pinyin converted from these dialects will increase. At least one first database can be screened out from the language database based on the length of the first pinyin, which can speed up the matching speed.

[0065] In addition, in the present invention, different keywords are set according to different dialects. For example, in "Xiao Fan, please start the voice block mode", "please" can be used as a keyword. In Cantonese, the pinyin of "please" is "cing", while in Mandarin, the pinyin of "please" is "qing". At this time, a part of the language database can be further screened out, and the remaining first database that meets the conditions will be marked as the second database. The matching speed is further accelerated, thereby greatly improving the user's operation convenience and satisfaction.

[0066] Finally, the first pinyin is matched with the second pinyin, and the second database with the highest matching value is used as the target language database. When matching, the second pinyin can be compared syllable by syllable by syllable through the first pinyin, and the language database corresponding to the second pinyin with the most identical syllables is used as the target language database.

[0067] Preferably, the steps of using the first pinyin to match the second pinyin are as follows:

[0068] The first pinyin and the second pinyin are encoded by one-hot, and vectorized to obtain a first vector and a second vector respectively;

[0069] The similarity between the first vector and the second vector is calculated, and the second vector with the highest similarity is obtained, so that the corresponding second database is used as the target language database.

[0070] Due to the syllable-by-syllable comparison method, the processing speed is slow. It is difficult to respond to users quickly, which affects the user experience. For this reason, in the present invention, the pinyin of the wake-up command in the language database can be first converted through one-hot to obtain a second vector, and at this time the second vector is stored in the cloud, which will not affect the storage memory of the local air conditioner. In subsequent operations, only the first pinyin needs to be processed, and after obtaining the first vector, the second vector is immediately called for calculation. Speed ​​up the response speed.

[0071] It is worth mentioning that in the calculation of vector similarity, the existing cosine similarity algorithm or Euclidean distance algorithm can be used to immediately calculate the result, which will not be demonstrated in detail in the present invention.

[0072] Preferably, the method further comprises the following steps:

[0073] Get the time when the language control configuration file is loaded as the first time;

[0074] Adding the threshold time to the first time to obtain a second time;

[0075] It is determined whether the language control configuration file is called within the second time. If it is called, the calling time is obtained and updated to the first time; if it is not called, the language control configuration file in the air conditioner is deleted.

[0076] The language control of a dialect requires downloading a corresponding language control configuration file. In actual testing, it is found that when the storage memory of the local air conditioner is fully loaded and the language control configuration file has not been completely downloaded into the air conditioner, the air conditioner and the cloud will be in a state of being connected all the time. After operating for a period of time, the processor of the air conditioner will heat up seriously, which will eventually affect the control of the air conditioner. For this reason, in the present invention, the language control configuration file corresponding to the dialect that has not been used for a long time will also be deleted to ensure that the storage memory space in the air conditioner can be completely downloaded to the language control configuration file. For example, when the language control configuration file is loaded at 16:01 on June 10, 2024, and the first threshold time is 15 days, the second time is 16:01 on June 25, 2024. If the language control profile is still not called during 16:01 on June 25, 2024, the language control profile of the dialect will be deleted. If the user uses the dialect to control the air conditioner at 16:01 on June 15, 2024, the language control profile of the dialect will be called. At this time, 16:01 on June 15, 2024 will be updated to the first time, and the second time will be confirmed by the first time. At this time, the second time will also be updated to 16:01 on June 130, 2024.

[0077] An air conditioning control system based on accented speech recognition, using the air conditioning control method based on accented speech recognition, comprising a setting module, an audio acquisition module and a cloud module;

[0078] The setting module is used to set language adjustment commands, and to construct a number of language databases for the language adjustment commands and various dialects;

[0079] The audio acquisition module is used to acquire the audio of the wake-up command input by the user and upload the audio to the cloud;

[0080] The cloud module is used to convert the audio into a first pinyin, find a target language database according to the first pinyin, associate a language control configuration file with a dialect in the target language database, and load the language control configuration file into the air conditioner.

[0081] Preferably, the cloud module includes a conversion submodule;

[0082] The conversion submodule is used to pre-process the audio of the input wake-up command to obtain processed data;

[0083] The MFCC speech feature sequence is extracted from the processed data to obtain the speech feature sequence, and the speech feature sequence is input into the existing acoustic layer recognition module. The recognition result in the form of the first pinyin grid is generated through the search algorithm of the acoustic layer recognition module.

[0084] Preferably, the cloud module includes a matching submodule;

[0085] The matching submodule is used to filter out at least one first database from the language database according to the length of the first pinyin;

[0086] Filter out at least one second-category database from the first database according to the key field in the first pinyin;

[0087] The first pinyin is used to match the second pinyins of the language adjustment commands in multiple second databases respectively, and the second database with the highest matching value is used as the target language database.

[0088] Preferably, the matching submodule includes a calculation unit;

[0089] The calculation unit is used to encode the first pinyin and the second pinyin respectively through one-hot, and vectorize them to obtain a first vector and a second vector respectively;

[0090] The similarity between the first vector and the second vector is calculated, and the second vector with the highest similarity is obtained, so that the corresponding second database is used as the target language database.

[0091] Preferably, it also includes a clearing module;

[0092] The clearing module is used to obtain the time when the language control configuration file is loaded as the first time;

[0093] Adding the threshold time to the first time to obtain a second time;

[0094] It is determined whether the language control configuration file is called within the second time. If it is called, the calling time is obtained and updated to the first time; if it is not called, the language control configuration file in the air conditioner is deleted.

[0095] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0096] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

Claims

1. An air conditioning control method based on accented speech recognition, characterized in that: The following steps are involved: Setting a language adjustment command, and constructing a number of language databases for the language adjustment command and various dialects; The language adjustment command includes a wake-up command and a control command; Acquire the audio of the wake-up command input by the user, and upload the audio to the cloud; The audio is converted into a first pinyin, a target language database is found from a language database according to the first pinyin, a language control configuration file associated with a dialect in the target language database is generated, and the language control configuration file is loaded into the air conditioner.

2. The air conditioning control method based on accented speech recognition according to claim 1, characterized in that: The steps of converting the audio into the first pinyin are as follows: Preprocessing the audio of the input wake-up command to obtain processed data; The MFCC speech feature sequence is extracted from the processed data to obtain the speech feature sequence, and the speech feature sequence is input into the existing acoustic layer recognition module. The recognition result in the form of the first pinyin grid is generated through the search algorithm of the acoustic layer recognition module.

3. The air conditioning control method based on accented speech recognition according to claim 1, characterized in that: The step of finding the target language database according to the first pinyin is as follows: Filtering at least one first database from the language database according to the length of the first pinyin; Filtering at least one second database from the first database according to the key field in the first pinyin; The first pinyin is used to match the second pinyin of the wake-up command in each second database, and the second database with the highest matching value is used as the target language database.

4. The air conditioning control method based on accented speech recognition according to claim 3, characterized in that: The steps to match the first pinyin with the second pinyin are as follows: The first pinyin and the second pinyin are encoded by one-hot, and vectorized to obtain a first vector and a second vector respectively; The similarity between the first vector and the second vector is calculated, and the second vector with the highest similarity is obtained, so that the corresponding second database is used as the target language database.

5. The air conditioning control method based on accented speech recognition according to claim 3, characterized in that: It also includes the following steps, Get the time when the language control configuration file is loaded as the first time; Adding the threshold time to the first time to obtain a second time; Determine whether the language control configuration file is called within the second time, and if called, obtain the calling time and update it to the first time; If it is not called, the language control configuration file in the air conditioner is deleted.

6. An air conditioning control system based on accented speech recognition, using the air conditioning control method based on accented speech recognition according to any one of claims 1 to 5, characterized in that: Including setting module, audio acquisition module and cloud module; The setting module is used to set language adjustment commands, and to construct a number of language databases for the language adjustment commands and various dialects; The audio acquisition module is used to acquire the audio of the wake-up command input by the user and upload the audio to the cloud; The cloud module is used to convert the audio into a first pinyin, find a target language database according to the first pinyin, associate a language control configuration file with a dialect in the target language database, and load the language control configuration file into the air conditioner.

7. The air conditioning control system based on accented speech recognition according to claim 6, characterized in that: The cloud module includes a conversion submodule; The conversion submodule is used to pre-process the audio of the input wake-up command to obtain processed data; The MFCC speech feature sequence is extracted from the processed data to obtain the speech feature sequence, and the speech feature sequence is input into the existing acoustic layer recognition module. The recognition result in the form of the first pinyin grid is generated through the search algorithm of the acoustic layer recognition module.

8. The air conditioning control system based on accented speech recognition according to claim 6, characterized in that: The cloud module includes a matching submodule; The matching submodule is used to filter out at least one first database from the language database according to the length of the first pinyin; Filter out at least one second-category database from the first database according to the key field in the first pinyin; The first pinyin is used to match the second pinyins of the language adjustment commands in multiple second databases respectively, and the second database with the highest matching value is used as the target language database.

9. The air conditioning control system based on accented speech recognition according to claim 8, characterized in that: The matching submodule includes a calculation unit; The calculation unit is used to encode the first pinyin and the second pinyin respectively through one-hot, and vectorize them to obtain a first vector and a second vector respectively; The similarity between the first vector and the second vector is calculated, and the second vector with the highest similarity is obtained, so that the corresponding second database is used as the target language database.

10. The air conditioning control system based on accented speech recognition according to claim 6, characterized in that: Also includes a clearing module; The clearing module is used to obtain the time when the language control configuration file is loaded as the first time; Adding the threshold time to the first time to obtain a second time; Determine whether the language control configuration file is called within the second time, and if called, obtain the calling time and update it to the first time; If it is not called, the language control configuration file in the air conditioner is deleted.

Images