Air conditioner self-adaptive adjusting method and device, air conditioner and storage medium

Abstract

The invention discloses an air conditioner self-adaptive adjusting method and device, an air conditioner and a storage medium. The method comprises the steps that indoor voice information is collected, and feature extraction is conducted on the voice information; classifying the voice information according to a feature extraction result to obtain a voice classification result; judging whether human voice information and / or noise information exists in the voice information based on the voice classification result; if it is judged that the voice information and / or the noise information exist in the voice information, generating a parameter adjustment instruction according to the voice information and / or the noise information; and the air conditioner is subjected to self-adaptive adjustment according to the parameter adjustment instruction. The operation process of the air conditioner is adjusted by collecting the environment sound, so that the problem that normal communication of users is disturbed due to too large noise when the air conditioner is in the operation states of music playing, high wind speed and the like can be solved, the purposes of reducing noise interference and improving the quality of the communication environment of the users are achieved, and the using effect of the air conditioner is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of air conditioners, and particularly to an air conditioner adaptive adjustment method, device, air conditioner and storage medium. Background Art

[0002] Currently, with the rapid development of smart home and Internet of Things technologies, the household air conditioning system is evolving deeply from the traditional single temperature control mode towards the intelligent and personalized direction. In recent years, the wide application of speech recognition technology in smart devices has enabled air conditioners to not only have basic functions such as remote control and timed switching, but also derive diversified and user-friendly services such as voice control and music playback, significantly improving the convenience of user operation. However, in actual application scenarios, when the air conditioner plays music, the background music in a complex environment is likely to become interference noise for normal conversations; and when the air conditioner operates at a high wind speed or in a strong gear, the generated wind noise will also have an adverse impact on indoor communication. Currently, when traditional air conditioning systems implement voice control and music playback functions, they mostly adopt a single control strategy, lacking a real-time monitoring mechanism for environmental noise and an adaptive adjustment means, and it is difficult to meet the user's demand for a quiet communication environment while ensuring a comfortable cooling (heating) effect. Summary of the Invention

[0003] Embodiments of the present invention provide an air conditioner adaptive adjustment method, device, air conditioner and storage medium, aiming to improve the use effect of the air conditioner.

[0004] In a first aspect, embodiments of the present invention provide an air conditioner adaptive adjustment method, including:

[0005] Collect voice information in the room and extract features from the voice information;

[0006] Classify the voice information according to the result of feature extraction to obtain a voice classification result;

[0007] Based on the voice classification result, determine whether there is human voice information and / or noise information in the voice information;

[0008] If it is determined that there is human voice information and / or noise information in the voice information, generate a parameter adjustment instruction according to the human voice information and / or noise information;

[0009] Perform adaptive adjustment on the air conditioner according to the parameter adjustment instruction.

[0010] In a second aspect, embodiments of the present invention provide an air conditioner adaptive adjustment device, including:

[0011] A voice collection unit, configured to collect voice information in the room and extract features from the voice information;

[0012] A voice classification unit for classifying voice information according to the result of feature extraction to obtain a voice classification result;

[0013] An information judgment unit for judging whether there is human voice information and / or noise information in the voice information based on the voice classification result;

[0014] An instruction generation unit for generating a parameter adjustment instruction according to the human voice information and / or noise information if it is determined that there is human voice information and / or noise information in the voice information;

[0015] An adaptive adjustment unit for adaptively adjusting the air conditioner according to the parameter adjustment instruction.

[0016] In a third aspect, an embodiment of the present invention provides an air conditioner, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the air conditioner adaptive adjustment method described in the first aspect is implemented.

[0017] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium with a computer program stored thereon. When the computer program is executed by a processor, the air conditioner adaptive adjustment method described in the first aspect is implemented.

[0018] An embodiment of the present invention provides an air conditioner adaptive adjustment method, device, air conditioner, and storage medium. The method includes: collecting voice information in a room and performing feature extraction on the voice information; classifying the voice information according to the result of feature extraction to obtain a voice classification result; judging whether there is human voice information and / or noise information in the voice information based on the voice classification result; if it is determined that there is human voice information and / or noise information in the voice information, generating a parameter adjustment instruction according to the human voice information and / or noise information; and adaptively adjusting the air conditioner according to the parameter adjustment instruction. Based on voice recognition and environment detection technologies, the embodiment of the present invention adjusts and sets the operating parameters of the air conditioner by collecting environmental sounds in real time and identifying conversation information, so as to solve the problem that current smart air conditioners interfere with normal user communication due to excessive noise during music playback and high wind speed operation, achieving the purpose of reducing noise interference and improving the quality of the user communication environment, thereby improving the use effect of the air conditioner. Description of the Drawings

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0020] Figure 1 Schematic flow chart of an air conditioner adaptive adjustment method provided by an embodiment of the present invention;

[0021] Figure 2 Schematic sub - flow chart of step S101 in an air conditioner adaptive adjustment method provided by an embodiment of the present invention;

[0022] Figure 3 Schematic sub - flow chart of step S102 in an air conditioner adaptive adjustment method provided by an embodiment of the present invention;

[0023] Figure 4 Schematic sub - flow chart of step S103 in an air conditioner adaptive adjustment method provided by an embodiment of the present invention;

[0024] Figure 5 Another schematic sub - flow chart of step S103 in an air conditioner adaptive adjustment method provided by an embodiment of the present invention;

[0025] Figure 6 Schematic block diagram of an air conditioner adaptive adjustment device provided by an embodiment of the present invention;

[0026] Figure 7 First schematic sub - block diagram of an air conditioner adaptive adjustment device provided by an embodiment of the present invention;

[0027] Figure 8 Second schematic sub - block diagram of an air conditioner adaptive adjustment device provided by an embodiment of the present invention;

[0028] Figure 9 Third schematic sub - block diagram of an air conditioner adaptive adjustment device provided by an embodiment of the present invention;

[0029] Figure 10 Fourth schematic sub - block diagram of an air conditioner adaptive adjustment device provided by an embodiment of the present invention;

[0030] Figure 11 Schematic block diagram of an air conditioner provided by an embodiment of the present invention. Detailed implementation manners

[0031] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0032] It should be understood that when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their combinations.

[0033] It should also be understood that the terms used in this specification of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. As used in this specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0034] It should be further understood that the term "and / or" used in this specification of the present invention and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0035] Please refer to the following Figure 1 , an embodiment of the present invention provides an air conditioner adaptive adjustment method, specifically including: steps S101 to S105.

[0036] Step S101, collect voice information in the room and extract features from the voice information;

[0037] Step S102, classify the voice information according to the result of feature extraction to obtain a voice classification result;

[0038] Step S103, based on the voice classification result, judge whether there is human voice information and / or noise information in the voice information;

[0039] Step S104, if it is determined that there is human voice information and / or noise information in the voice information, generate a parameter adjustment instruction according to the human voice information and / or noise information;

[0040] Step S105, perform adaptive adjustment on the air conditioner according to the parameter adjustment instruction.

[0041] In this embodiment, first, the voice information in the room is collected, such as noise, voices, and other sound sources. Feature extraction is performed on these sound sources to achieve voice classification, that is, to separate sound sources such as noise and human voices from the voice information. Then, parameter adjustment instructions for the air conditioner are generated based on the human voice information and / or noise information therein, and the operation parameters of the air conditioner are automatically adjusted according to the parameter adjustment instructions. This embodiment is based on voice recognition and environment detection technologies. By collecting environmental sounds in real time and identifying conversation information, the operation parameters of the air conditioner are adjusted and set. In this way, the problem that current smart air conditioners interfere with normal user communication due to excessive noise during music playback and high-speed operation can be solved, achieving the purpose of reducing noise interference and improving the quality of the user's communication environment, thereby enhancing the usage effect of the air conditioner.

[0042] In an actual application scenario, when the air conditioner is in the music playback mode, the air conditioner adaptive adjustment method provided in this embodiment can generate parameter adjustment instructions based on whether there is human voice information in the room, and adjust the music volume through the parameter adjustment instructions to avoid affecting normal user conversation and provide a better communication environment for users. Another example is when the air conditioner is operating at high speed and there is human voice information in the room. It is possible to determine whether to adjust the wind speed based on the wind noise generated by the high speed to reduce the interference of the wind noise on the human voice in the room.

[0043] In one embodiment, as Figure 2 shown, step S101 includes steps S201 to S208.

[0044] Step S201, preprocess the voice information; wherein, the preprocessing includes noise reduction, filtering, and gain adjustment processing;

[0045] Step S202, apply a pre-emphasis filter to the preprocessed voice information to perform pre-emphasis processing on the voice information;

[0046] Step S203, perform frame segmentation processing on the pre-emphasized voice information to obtain multiple frames of voice signals;

[0047] Step S204, for each frame of voice signal, add a preset window function and perform a fast Fourier transform on the voice signal to obtain corresponding spectral information;

[0048] Step S205, based on the spectral information, calculate the square of the amplitude of the fast Fourier transform result to obtain the corresponding power spectrum;

[0049] Step S206, use a Mel filter bank to convert the power spectrum from a linear frequency scale to a Mel frequency scale to obtain an energy spectrum;

[0050] Step S207: Take the logarithm of the energy spectrum and perform a discrete cosine transform on the logarithmized energy spectrum;

[0051] Step S208: Extract the first n coefficients from the result of the discrete cosine transform as the MFCC feature vector, and use the MFCC feature vector as the result of the feature extraction of the speech information.

[0052] In this embodiment, when extracting features from speech information, first perform preprocessing on it. For example, perform noise reduction processing (digital filtering algorithms such as adaptive filtering and frequency-domain filtering can be used) to eliminate irrelevant noise in the environment, and perform gain adjustment and other processing to ensure the clarity of the original speech data. Then, extract feature vectors such as MFCC (Mel Frequency Cepstral Coefficients) from the speech information for subsequent input into a deep learning model for speech classification processing. Specifically, first apply a pre-emphasis filter to the speech information to compensate for the insufficient energy in the high-frequency part of the speech signal. Then divide the pre-emphasized signal into frames at 20 milliseconds or other values, and use a certain frame shift of 10 milliseconds to ensure that the signal within each frame is approximately stationary. Next, multiply each frame of the signal by a window function to reduce the spectral leakage problem at the frame edges. Then perform a fast Fourier transform (FFT) on each windowed frame of the signal to convert the speech signal to the frequency domain and obtain the spectral information of this frame. Then calculate the squared magnitude of the FFT result to obtain the power spectrum of this frame. Subsequently, pass the power spectrum through a set of triangular Mel filters, and the number of filters can be set to 20, converting the linear frequency scale to a Mel frequency scale that is more in line with the human ear's perception. Take the logarithm of the energy values after passing through the Mel filters to simulate the logarithmic response characteristic of the human ear to loudness. Finally, perform a DCT transform on the logarithmized Mel energy spectrum and extract the first n (for example, the first 12) coefficients, and these coefficients are the MFCC feature vectors and can be used for subsequent recognition processing.

[0053] In one embodiment, as Figure 3 shown, the step S102 includes: steps S301 to S305.

[0054] Step S301: Construct a two-dimensional feature matrix based on the result of feature extraction and input the two-dimensional feature matrix into a multi-layer convolutional neural network for convolutional operations;

[0055] Step S302: Extract local time-frequency domain features through convolutional kernels of different sizes to capture local features;

[0056] Step S303: Use the pooling layer to perform pooling processing on the local features to obtain a feature sequence;

[0057] Step S304: Input the feature sequence into a recurrent neural network, capture time dynamics by combining with a long short-term memory network, and then use the hidden state output by the recurrent neural network as high-level semantic features for feature fusion;

[0058] Step S305: Input the result of feature fusion into a fully connected layer, and perform multi-class classification by combining with a softmax layer to obtain a voice classification result.

[0059] In this embodiment, when classifying the extracted features, first, the extracted features, such as MFCC feature vectors, are constructed into a two-dimensional feature matrix. In this two-dimensional feature matrix, the horizontal axis can represent time frames, and the vertical axis can represent each MFCC coefficient. Subsequently, a multi-layer convolutional neural network (CNN) is used to perform convolutional operations on the two-dimensional feature matrix, and local time-frequency domain features are extracted through convolutional kernels of different sizes to capture short-term changes and local texture information in the speech signal. Usually, a pooling layer is introduced after the convolutional layer to further extract robust features and reduce the dimension, and a feature sequence is obtained. Then, the feature sequence processed by the CNN is input into a recurrent neural network (RNN), and a long short-term memory network (LSTM) is used to capture long-term dependence characteristics. Here, the temporal dynamic changes of the speech signal can be modeled according to the context relationship to distinguish the temporal patterns of human voices, background music, and wind noise in the speech. The hidden state of the RNN is used as high-level semantic features for feature fusion. Then, the feature fusion result output by the RNN is mapped to the feature space through a fully connected layer, and a softmax layer is used for multi-class classification. For example, the probabilities of multiple classes are output, such as human voices, background music, and other noise types. Further, the probability output result and continuous frames can be combined for comprehensive judgment to improve the stability and accuracy of voice classification.

[0060] In one embodiment, as Figure 4 shown, step S103 includes steps S401 to S405.

[0061] Step S401: When it is determined that there is human voice information in the voice message, extract the human voice information from the voice message;

[0062] Step S402: Use the VAD algorithm to distinguish the speech part and the non-speech part of the human voice information, and extract the continuous speech segments in the human voice information according to the discrimination result;

[0063] Step S403: Use a speaker separation algorithm to perform speaker segmentation on the continuous speech segments to obtain speaker embedding vectors;

[0064] Step S404: Use a clustering algorithm to perform clustering discrimination on the speaker embedding vectors to obtain the number of speakers;

[0065] Step S405: Determine whether it is in a conversation state based on the number of speakers, a preset time threshold, and a preset sound intensity threshold, and generate a conversation signal when it is determined that it is in a conversation state.

[0066] After obtaining the voice classification result, if there is voice information therein, it is possible to further determine whether it is in a conversation state based on the voice information, that is, to determine whether there are multiple people having a conversation. If so, the operating parameters of the air conditioner can be adjusted to reduce interference such as background music interference or noise. Specifically, first, the VAD (Voice Activity Detection) algorithm is used to distinguish the voice part and the non-voice part in the voice information, and the continuous voice segments are segmented. Then, the speaker diarization algorithm is used for the continuous voice segments to obtain speaker embedding vectors. Next, a clustering algorithm is used for the speaker embedding vectors extracted from each voice segment to distinguish different speakers. Here, the number of different clusters in the clustering result is the number of detected speakers. When the number of detected different speaker categories is greater than or equal to 2, it can be determined that it is a conversation state of two or more people, and a conversation signal is generated accordingly. Further, analyze the change of speakers in a continuous time period. If there are multiple speaker switches within a certain time window and the clustering result shows at least two categories or more, the continuity of the conversation is confirmed. It is also possible to set a specific time window and a speaker change threshold, such as the length of the conversation segment exceeding a certain time, the number of speaker switches reaching a preset value, etc., to further improve the accuracy of the determination. At the same time, the volume and signal strength information can also be referred to ensure that the detected conversation segment is a real and effective communication signal. For example, a threshold T_voice is set. When the characteristics of continuous conversation of two or more people are detected and the signal strength exceeds T_voice, it is determined that there is a conversation in the room, and a conversation signal is generated.

[0067] In one embodiment, as Figure 5 shown, the step S103 further includes: steps S501 to S503.

[0068] Step S501: Use a wind speed sensor to collect the wind noise information of the air conditioner, and set a noise mean value and extract a noise peak value in combination with the wind noise information and the noise information;

[0069] Step S502: Compare the noise mean value with a preset first noise threshold, and compare the noise peak value with a preset second noise threshold;

[0070] Step S503: If the noise mean value reaches the preset first noise threshold, and / or the noise peak value reaches the preset second noise threshold, generate a noise signal.

[0071] The wind noise information generated during the operation of the air conditioner is collected by a wind speed sensor, and the noise information extracted by voice classification is combined to set the noise mean value, and the noise peak value is determined therefrom. Then, the noise mean value and the noise peak value are respectively compared with a preset noise threshold value to determine whether the noise mean value and the noise peak value reach a certain level. If any one of the noise mean value and the noise peak value reaches the preset noise threshold value, a noise signal is generated for subsequent operation adjustment through the noise signal. For example, the noise mean value is N_avg, the noise peak value is N_peak, and the preset noise threshold value N_threshold = 50 dB. When it is detected that N_avg = 65 dB, at this time N_avg > N_threshold, and a human voice dialogue signal is detected, subsequent parameter adjustment is triggered.

[0072] In an actual application scenario, the noise information in the room is collected by a noise sensor, and the wind noise information of the air conditioner is collected by a wind speed sensor. Among them, the noise sensor can be independent of the air conditioner. For example, the noise sensor can be configured in a sub-module of the environmental intelligent detection device, and the environmental intelligent detection device can be used as a separate intelligent product to be linked with the air conditioner device. In this way, the environmental intelligent detection device can detect the noise of the indoor environment through the noise sensor and then send the data to the air conditioner for the air conditioner to process the data. At the same time, the wind speed sensor can be configured at a relevant position of the air outlet of the air conditioner to detect the wind noise data, so that the noise generated during the air supply process of the air conditioner can be collected by the wind speed sensor. For example, there will be a large wind noise when the air conditioner is in the high wind gear. Of course, the noise sensor can also be integrated into the air conditioner to be responsible for detecting the noise of the indoor environment, while the wind speed sensor is configured at a relevant position of the air outlet of the air conditioner to be responsible for detecting the noise value of the air outlet.

[0073] In one embodiment, the step S104 includes:

[0074] Obtain the operating mode of the air conditioner;

[0075] When the air conditioner is in the music playing mode, if there is the dialogue signal, generate a volume adjustment instruction according to the following formula:

[0076] V_new = V_original × (1 - α);

[0077] wherein, V_new represents the volume to be adjusted, V_original represents the current volume, and α represents the adjustment coefficient;

[0078] When the air conditioner is in the high wind speed mode, if there are both the dialogue signal and the noise signal at the same time, generate a wind speed adjustment instruction according to the following formula:

[0079] S_new = S_original - ΔS;

[0080] Wherein, S_new represents the wind speed to be adjusted, S_original represents the current wind speed, and ΔS represents the wind speed adjustment step size.

[0081] In this embodiment, when generating a parameter adjustment instruction, a decision tree algorithm can be used to analyze the detection data, and the decision logic is divided into two modes:

[0082] One is in the music playback mode. If a dialogue signal is detected, the following volume adjustment formula is executed:

[0083] V_new = V_original × (1 - α);

[0084] Wherein, α is an adjustment coefficient. In practical applications, it can take a value of 0.3 and can be dynamically determined by subsequent self-learning of the current noise intensity.

[0085] The other is in the high wind speed mode. If both a dialogue signal and a noise signal are detected, the following wind speed adjustment formula is executed:

[0086] S_new = S_original - ΔS;

[0087] Wherein, ΔS is the wind speed adjustment step size, which can be specifically determined by the performance parameters of the air conditioning equipment.

[0088] This embodiment organically combines speech recognition, environmental noise detection, and air conditioner operation parameter automatic adjustment technologies, and can achieve that when the air conditioner plays music and detects the presence of two or more people's conversations, the speaker volume is automatically reduced to avoid interference of background music on communication; and when the air conditioner operates at high wind speed or strong gear, when detecting conversations indoors, the wind speed parameters are automatically adjusted to reduce wind noise while ensuring the air conditioning cooling (heating) effect.

[0089] In practical applications, in addition to generating a wind speed adjustment instruction according to the dialogue signal and the noise signal, a wind speed adjustment instruction can also be generated solely based on the noise signal. For example, when the noise mean or noise peak reaches another set level, it can be determined that the current noise information will have a certain impact on the normal life of the user. Therefore, a noise signal can be generated. Even if there is no dialogue signal at this time, a wind speed adjustment instruction can still be generated to reduce the impact of the noise information on the user.

[0090] In one embodiment, the adaptive adjustment of the air conditioner further includes:

[0091] Responding to an intervention adjustment instruction issued by the user, and performing adaptive adjustment on the air conditioner according to the intervention adjustment instruction.

[0092] In addition to the air conditioner automatically adjusting its operating parameters according to the acquired data, the user can also send intervention instructions to the air conditioner through a control terminal such as an APP or a remote control, so that the air conditioner adjusts the operating parameters according to the instructions sent by the user. For example, in a home or office environment, after the user starts the air conditioner music playback through a voice command, the air conditioner collects voice information and extracts the dialogue signal and noise signal from it. At this time, if the air conditioner is in the music playback mode, it automatically lowers the speaker volume to ensure clear dialogue content. In addition, when the user sets the air conditioner to the high wind speed mode for rapid cooling, if the air conditioner detects that there are multiple people talking indoors, it reduces the fan speed according to the wind speed adjustment formula to reduce wind noise while ensuring indoor air circulation.

[0093] Preferably, the user can also use terminal tools such as a mobile APP to view the adaptive adjustment curve and parameter changes of the air conditioner in real time and make manual fine-tuning according to personal needs. In this way, through long-term operation and data feedback, the parameter settings are continuously optimized to achieve preventive noise control, enabling the air conditioner to intelligently match the environmental requirements in any usage scenario, ensuring both comfort and reducing noise interference, and comprehensively improving the quality of the living and working environments.

[0094] In the actual application scenario, the adjustment instructions sent by the user or the adjustment instructions generated by the air conditioner can be transmitted to the main control module of the air conditioner through a communication protocol. This module can activate the speaker and fan drive systems to achieve real-time control of each component of the air conditioner. After the adjustment is completed, the air conditioner can feedback the current operating status (such as volume level, wind speed level, and noise detection value) to the user through the user interaction module. The user can view the historical adjustment situation through the display screen or the mobile APP and can intervene and adjust the historical data by himself for better adaptive adjustment of this scenario in the future (such as adjusting ΔS).

[0095] Correspondingly, in addition to the main control module, a voice collection module unit, a voice recognition module unit, an environment detection module unit, and a parameter adjustment control module unit can also be set. Among them, the voice collection module can adopt a high-sensitivity microphone unit to collect the sound signals of the indoor environment in real time. The voice recognition module can use deep learning algorithms (such as the combined model of convolutional neural network CNN and recurrent neural network RNN) to perform operations such as frame segmentation, feature extraction, and pattern recognition on the collected voice signals. The environment detection module can be equipped with a noise sensor to monitor the indoor environmental noise in real time, and the data is fed back in decibels (dB). At the same time, a wind speed sensor is configured to detect the wind speed and wind noise during the operation of the air conditioner. The parameter adjustment control module is used to receive the data from the voice recognition module and the environment detection module, and adopt a decision tree algorithm control strategy to judge and adjust the current air conditioner operation parameters. The main control module of the air conditioner is used to receive the parameter adjustment instructions and drive each subsystem inside the air conditioner (including speakers, fans, and other control units) to adjust the operation state.

[0096] In addition, during the operation of the air conditioner, all the collected data, adjustment records, and user feedback can be stored in the local database and regularly uploaded to the cloud server. The cloud data center uses big data analysis and machine learning algorithms to summarize and predict the historical data, forming a dynamic adjustment parameter library. The air conditioner can continuously update the voice recognition model and control strategy according to different usage scenarios (such as home, office, etc.) to achieve adaptive learning and online upgrade, so as to always maintain the optimal adjustment effect.

[0097] Figure 6 FIG. 600 is a schematic block diagram of an air conditioner adaptive adjustment device 600 provided by an embodiment of the present invention. The device 600 includes:

[0098] A voice collection unit 601, configured to collect voice information in the room and extract features from the voice information;

[0099] A voice classification unit 602, configured to classify the voice information according to the result of feature extraction to obtain a voice classification result;

[0100] An information judgment unit 603, configured to judge whether there is human voice information and / or noise information in the voice information based on the voice classification result;

[0101] An instruction generation unit 604, configured to generate a parameter adjustment instruction according to the human voice information and / or noise information if it is determined that there is human voice information and / or noise information in the voice information;

[0102] An adaptive adjustment unit 605, configured to perform adaptive adjustment on the air conditioner according to the parameter adjustment instruction.

[0103] In one embodiment, asFigure 7 As shown, the voice collection unit 601 includes:

[0104] A preprocessing unit 701 for preprocessing the voice information; wherein, the preprocessing includes noise reduction, filtering and gain adjustment processing;

[0105] A pre-emphasis unit 702 for applying a pre-emphasis filter to the preprocessed voice information to perform pre-emphasis processing on the voice information;

[0106] A frame segmentation processing unit 703 for performing frame segmentation processing on the pre-emphasized voice information to obtain multiple frames of voice signals;

[0107] A first transformation unit 704 for adding a preset window function to each frame of voice signal and performing a fast Fourier transform on the voice signal to obtain corresponding spectrum information;

[0108] An amplitude calculation unit 705 for calculating the square of the amplitude of the fast Fourier transform result based on the spectrum information to obtain a corresponding power spectrum;

[0109] A scale transformation unit 706 for converting the power spectrum from a linear frequency scale to a Mel frequency scale by using a Mel filter bank to obtain an energy spectrum;

[0110] A second transformation unit 707 for taking the logarithm of the energy spectrum and performing a discrete cosine transform on the logarithmized energy spectrum;

[0111] A vector extraction unit 708 for extracting the first n coefficients in the result of the discrete cosine transform as an MFCC feature vector and using the MFCC feature vector as the result of the feature extraction of the voice information.

[0112] In one embodiment, as Figure 8 shown, the voice classification unit 602 includes:

[0113] A matrix construction unit 801 for constructing a two-dimensional feature matrix based on the result of feature extraction and inputting the two-dimensional feature matrix into a multi-layer convolutional neural network for convolutional operation;

[0114] A local capture unit 802 for extracting local time-frequency domain features through convolutional kernels of different sizes to capture local features;

[0115] A pooling processing unit 803 for performing pooling processing on the local features by using a pooling layer to obtain a feature sequence;

[0116] A feature fusion unit 804, configured to input the feature sequence into a recurrent neural network, perform time dynamics capture in combination with a long short-term memory network, and then perform feature fusion using the hidden state output by the recurrent neural network as a high-level semantic feature;

[0117] A category classification unit 805, configured to input the result of feature fusion into a fully connected layer, and perform multi-category classification in combination with a softmax layer to obtain a voice classification result.

[0118] In one embodiment, as Figure 9 shown, the information judgment unit 603 includes:

[0119] A human voice extraction unit 901, configured to extract the human voice information from the voice information when it is determined that there is human voice information in the voice information;

[0120] A voice discrimination unit 902, configured to use a VAD algorithm to distinguish the voice part and the non-voice part of the human voice information, and extract continuous voice segments in the human voice information according to the discrimination result;

[0121] A segmentation processing unit 903, configured to perform speaker segmentation processing on the continuous voice segments using a speaker separation algorithm to obtain speaker embedding vectors;

[0122] A clustering discrimination unit 904, configured to collect clustering algorithms to perform clustering discrimination on the speaker embedding vectors to obtain the number of speakers;

[0123] A first signal generation unit 905, configured to determine whether it is in a conversation state in combination with the number of speakers, a preset time threshold, and a preset sound intensity threshold, and generate a conversation signal when it is determined that it is in a conversation state.

[0124] In one embodiment, as Figure 10 shown, the information judgment unit 603 further includes:

[0125] A noise setting unit 1001, configured to collect wind noise information of the air conditioner using a wind speed sensor, and set a noise mean value and extract a noise peak value in combination with the wind noise information and the noise information;

[0126] A noise comparison unit 1002, configured to compare the noise mean value with a preset first noise threshold, and compare the noise peak value with a preset second noise threshold;

[0127] A second signal generation unit 1003, configured to generate a noise signal if the noise mean value reaches the preset first noise threshold, and / or the noise peak value reaches the preset second noise threshold.

[0128] In one embodiment, the instruction generation unit 604 includes:

[0129] A mode acquisition unit, configured to acquire the operating mode of the air conditioner;

[0130] A volume command generation unit, configured to, when the air conditioner is in the music playback mode, if there is the dialogue signal, generate a volume adjustment command according to the following formula:

[0131] V_new = V_original × (1 - α);

[0132] wherein, V_new represents the volume to be adjusted, V_original represents the current volume, and α represents the adjustment coefficient;

[0133] A wind speed command generation unit, configured to, when the air conditioner is in the high wind speed mode, if there are both the dialogue signal and the noise signal, generate a wind speed adjustment command according to the following formula:

[0134] S_new = S_original - ΔS;

[0135] wherein, S_new represents the wind speed to be adjusted, S_original represents the current wind speed, and ΔS represents the wind speed adjustment step size.

[0136] In an embodiment, the air conditioner adaptive adjustment device 600 further includes:

[0137] An intervention adjustment unit, configured to respond to an intervention adjustment command issued by a user, and perform adaptive adjustment on the air conditioner according to the intervention adjustment command.

[0138] Since the embodiments of the device part correspond to the embodiments of the method part, for the embodiments of the device part, please refer to the description of the embodiments of the method part, which will not be elaborated here for the time being.

[0139] Please refer to Figure 11 , Figure 11 which is a schematic block diagram of an air conditioner 1100 provided by an embodiment of the present invention. The air conditioner 1100 is a device with wireless communication and wired communication.

[0140] Refer to Figure 11 , the air conditioner 1100 includes a processor 1102, a memory, and a network interface 1105 connected through a system bus 1101, wherein the memory may include a non-volatile storage medium 1103 and an internal memory 1104.

[0141] The non-volatile storage medium 1103 can store an operating system 11031 and a computer program 11032. When the computer program 11032 is executed, the processor 1102 can be caused to execute an air conditioner control method.

[0142] The processor 1102 is used to provide computing and control capabilities to support the operation of the entire air conditioner 1100.

[0143] The internal memory 1104 provides an environment for the operation of the computer program 11032 in the non-volatile storage medium 1103. When the computer program 11032 is executed by the processor 1102, the processor 1102 can be caused to execute an air conditioner control method.

[0144] The network interface 1105 is used for network communication with other devices. Those skilled in the art can understand that Figure 11 the structure shown in is only a block diagram of some structures related to the solution of the present invention, and does not constitute a limitation on the air conditioner 1100 to which the solution of the present invention is applied. The specific air conditioner 1100 may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0145] Among them, the processor 1102 is used to run the computer program 11032 stored in the memory to implement any embodiment of the above-mentioned air conditioner control method.

[0146] It should be understood that in the embodiments of the present invention, the processor 1102 may be a central processing unit (CPU), and the processor 1102 may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0147] Those of ordinary skill in the art can understand that all or part of the processes in the methods of implementing the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a storage medium, and the storage medium is a computer-readable storage medium. The computer program is executed by at least one processor in the computer system to implement the process steps of the embodiments of the above method.

[0148] An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed, the steps provided in the above embodiments can be implemented. The storage medium may include: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc.

[0149] The various embodiments in the specification are described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple. For the relevant parts, reference can be made to the description in the method section. It should be noted that for those of ordinary skill in the art in the technical field of the present application, without departing from the principle of the present application, several improvements and modifications can be made to the present application, and these improvements and modifications also fall within the protection scope of the claims of the present application.

[0150] It should also be noted that in this specification, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, article or device including the said element.

Claims

1. An air conditioner adaptive adjustment method, characterized in that Including: Collecting voice information in the room and extracting features from the voice information; Classifying the voice information according to the results of feature extraction to obtain a voice classification result; Based on the voice classification result, determining whether there is human voice information and / or noise information in the voice information; If it is determined that there is human voice information and / or noise information in the voice information, generating a parameter adjustment instruction according to the human voice information and / or noise information; Performing adaptive adjustment on the air conditioner according to the parameter adjustment instruction.

2. The air conditioner adaptive adjustment method according to claim 1, characterized in that The collecting voice information in the room and extracting features from the voice information includes: Performing preprocessing on the voice information; wherein, the preprocessing includes noise reduction, filtering, and gain adjustment processing; Applying a pre-emphasis filter to the preprocessed voice information to perform pre-emphasis processing on the voice information; Performing frame segmentation on the pre-emphasized voice information to obtain multiple frames of voice signals; For each frame of voice signal, adding a preset window function and performing a fast Fourier transform on the voice signal to obtain corresponding spectrum information; Based on the spectrum information, calculating the square of the magnitude of the fast Fourier transform result to obtain a corresponding power spectrum; Using a Mel filter bank to convert the power spectrum from a linear frequency scale to a Mel frequency scale to obtain an energy spectrum; Taking the logarithm of the energy spectrum and performing a discrete cosine transform on the logarithmized energy spectrum; Extracting the first n coefficients in the result of the discrete cosine transform as MFCC feature vectors, and using the MFCC feature vectors as the results of feature extraction of the voice information.

3. The air conditioner adaptive adjustment method according to claim 1, wherein The classifying the voice information according to the results of feature extraction to obtain a voice classification result includes: Constructing a two-dimensional feature matrix based on the results of feature extraction and inputting the two-dimensional feature matrix into a multi-layer convolutional neural network for convolutional operations; Extracting local time-frequency domain features through convolutional kernels of different sizes to capture local features; Using a pooling layer to perform pooling processing on the local features to obtain a feature sequence; Inputting the feature sequence into a recurrent neural network, dynamically capturing time through a long short-term memory network, and then fusing the hidden states output by the recurrent neural network as high-level semantic features; Inputting the result of feature fusion into a fully connected layer and performing multi-class classification in combination with a softmax layer to obtain a voice classification result.

4. The air conditioner adaptive adjustment method according to claim 1, wherein The determining whether there is human voice information and / or noise information in the voice information based on the voice classification result includes: When it is determined that there is human voice information in the voice information, extracting the human voice information from the voice information; Using a VAD algorithm to distinguish the voice part and the non-voice part of the human voice information, and extracting continuous voice segments in the human voice information according to the discrimination result; Using a speaker separation algorithm to perform speaker segmentation processing on the continuous voice segments to obtain speaker embedding vectors; Collecting a clustering algorithm to perform clustering discrimination on the speaker embedding vectors to obtain the number of speakers; Combining the number of speakers, a preset time threshold, and a preset sound intensity threshold to determine whether it is in a conversation state, and generating a conversation signal when it is determined to be in a conversation state.

5. The air conditioner adaptive adjustment method according to claim 4, wherein, Based on the voice classification result, determining whether there is human voice information and / or noise information in the voice message further includes: Collecting the wind noise information of the air conditioner by using a wind speed sensor, and setting a noise mean value and extracting a noise peak value in combination with the wind noise information and the noise information; Comparing the noise mean value with a preset first noise threshold, and comparing the noise peak value with a preset second noise threshold; If the noise mean value reaches the preset first noise threshold, and / or the noise peak value reaches the preset second noise threshold, generating a noise signal.

6. The air conditioner adaptive adjustment method according to claim 5, wherein If it is determined that there is human voice information and / or noise information in the voice message, generating a parameter adjustment instruction according to the human voice information and / or the noise information includes: Obtaining the operating mode of the air conditioner; When the air conditioner is in the music playing mode, if there is the dialogue signal, generating a volume adjustment instruction according to the following formula: V_new = V_original×(1 - α); Wherein, V_new represents the volume to be adjusted, V_original represents the current volume, and α represents an adjustment coefficient; When the air conditioner is in the high wind speed mode, if there are both the dialogue signal and the noise signal, generating a wind speed adjustment instruction according to the following formula: S_new = S_original - ΔS; Wherein, S_new represents the wind speed to be adjusted, S_original represents the current wind speed, and ΔS represents a wind speed adjustment step length.

7. The air conditioner adaptive adjustment method according to claim 1, wherein, Further includes: Responding to an intervention adjustment instruction issued by a user, and performing adaptive adjustment on the air conditioner according to the intervention adjustment instruction.

8. An air conditioner adaptive adjustment device, characterized in that, Includes: A voice collection unit, configured to collect voice information in a room and extract features of the voice information; A voice classification unit, configured to classify the voice information according to the result of feature extraction to obtain a voice classification result; An information judgment unit, configured to determine whether there is human voice information and / or noise information in the voice message based on the voice classification result; An instruction generation unit, configured to generate a parameter adjustment instruction according to the human voice information and / or the noise information if it is determined that there is human voice information and / or noise information in the voice message; An adaptive adjustment unit, configured to perform adaptive adjustment on the air conditioner according to the parameter adjustment instruction.

9. An air conditioner, characterized in that, Includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the adaptive adjustment method of the air conditioner according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the adaptive adjustment method of the air conditioner according to any one of claims 1 to 7 is implemented.

Images