A rechargeable hair dryer and control method thereof

Abstract

The application provides a rechargeable hair dryer and a control method thereof, comprising: a voice recognition module, configured to receive user voice data, perform data processing, recognize the user voice data after the data processing, and determine a control instruction; the control instruction comprises a temperature control instruction and a wind power control instruction; a gear adjustment module, configured to receive the control instruction sent by the voice recognition module, determine a target working gear, and perform adjustment. The application aims to improve the convenience of using the hair dryer by the mode of charging driving, intuitively displaying the wind temperature and the wind power, and voice control adjusting the wind temperature and the wind power, so as to improve the satisfaction of users.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of hair dryers, in particular to a rechargeable hair dryer and a control method thereof. BACKGROUND

[0002] At present, as the most common small household appliances for hairdressing, the popularity rate of hair dryers among consumers in cities and towns in China is increasing year by year. It can be said that the hair dryer has become one of the essential small household appliances for urban residents. Not only has the popularity rate of hair dryers improved, but also the use frequency of users has improved.

[0003] With the increasing demand for hair dryers, more requirements for the quality and functions of hair dryers are put forward. The traditional hair dryer is driven by connecting a 220V alternating current through a power cord. Such a hair dryer is always bound by the power cord during use, which is inconvenient to use, not to mention in an environment away from the power supply.

[0004] In addition, during the use of the hair dryer, the user cannot intuitively know the temperature and wind power of the air outlet, and often needs to adjust the temperature and wind power of the air outlet several times according to the feeling. However, the traditional hair dryer needs to be manually operated several times to achieve the desired effect of the user, and the adjustment process is very tedious and inconvenient. SUMMARY

[0005] The present application aims to at least solve one of the above technical problems. To this end, the present application aims to provide a rechargeable hair dryer and a control method thereof, which aims to improve the convenience of using the hair dryer by charging, intuitively displaying the temperature and wind power of the air outlet, and adjusting the temperature and wind power of the air outlet by voice control, thereby improving the satisfaction of the user.

[0006] To achieve the above-mentioned purpose, the present application provides a rechargeable hair dryer, comprising:

[0007] A voice recognition module is configured to receive user voice data and perform data processing, recognize the user voice data after data processing, and determine a control instruction; the control instruction includes a temperature control instruction and a wind power control instruction;

[0008] A gear adjustment module is configured to receive the control instruction sent by the voice recognition module, determine a target working gear and adjust it.

[0009] According to some embodiments of the present application, the rechargeable hair dryer further comprises:

[0010] A rechargeable battery pack;

[0011] A battery management module is configured to monitor the storage and working states of the rechargeable battery pack and manage the rechargeable battery pack based on the storage and working states.

[0012] According to some embodiments of the present application, the voice recognition module comprises:

[0013] A first receiving sub-module is configured to receive user voice data.

[0014] A processing sub-module is configured to process the user voice data to obtain to-be-recognized user voice data.

[0015] A recognition sub-module is configured to perform voice recognition on the to-be-recognized user voice data to determine a control instruction.

[0016] According to some embodiments of the present application, the processing sub-module comprises:

[0017] A noise reduction processing unit is configured to reduce noise signals in the user voice data to obtain first intermediate voice data.

[0018] A voice separation unit is configured to separate the user voice data from the first intermediate voice data as second intermediate voice data.

[0019] A signal enhancement unit is configured to perform high-frequency signal enhancement processing on the second intermediate voice data to obtain to-be-recognized user voice data.

[0020] According to some embodiments of the present application, the recognition sub-module comprises:

[0021] A feature extraction unit is configured to input the to-be-recognized user voice data into a pre-trained voice feature extraction model to obtain feature vector information of the to-be-recognized user voice data as to-be-matched feature vector information.

[0022] A feature matching unit is configured to match the feature vector information with keyword voice feature vectors in a preset local voice library to obtain a matching result.

[0023] A semantic acquisition unit is configured to obtain first user voice semantics based on a semantic correspondence relationship between the matching result and the matched keyword voice feature vectors; the first user voice semantics comprise function description and adjustment metric description.

[0024] A judgment unit is configured to perform validity judgment on the first user voice semantics, take the first user voice semantics that pass the validity judgment as target user voice semantics, and determine a control instruction according to the target user voice semantics.

[0025] According to some embodiments of the present application, the judgment unit comprises:

[0026] The first determining subunit is configured to determine the function description in the first user voice semantics and determine the function that needs to be adjusted.

[0027] The second determining subunit is configured to determine the adjustment metric description in the first user voice semantics after the first determining subunit determines the function that needs to be adjusted, and determine the target value.

[0028] The third determining subunit is configured to:

[0029] determine the adjustment range corresponding to the function that needs to be adjusted;

[0030] when it is determined that the target value is within the adjustment range, it is determined that the effectiveness determination is passed, the first user voice semantics that passes the effectiveness determination is taken as the target user voice semantics, and the control instruction is determined according to the target user voice semantics.

[0031] According to some embodiments of the present application, the noise reduction processing unit comprises:

[0032] The voice screening processing subunit is configured to remove the pure silence data segment in the user voice data to obtain noise user voice data.

[0033] The voice data processing subunit is configured to determine the noise frame and the voice frame in the noise user voice data.

[0034] The model training subunit is configured to train a target noise reduction neural network model based on a training model sample.

[0035] The user voice noise reduction subunit is configured to obtain user noise reduction voice data based on the voice frame and the target noise reduction neural network model, and take the user noise reduction voice data as the first intermediate voice data.

[0036] According to some embodiments of the present application, the voice separation unit comprises:

[0037] The training subunit is configured to train a voice segmentation model based on sample original voice data and corresponding sample final voice data.

[0038] The separation subunit is configured to:

[0039] obtain a segmented voice embedding feature sequence set of the first intermediate voice data based on the first intermediate voice data and the trained voice segmentation model;

[0040] perform voice clustering on the first intermediate voice data based on the segmented voice embedding feature sequence set to obtain voice data of the user in the first intermediate voice data, and take the voice data as second intermediate voice data.

[0041] According to some embodiments of the present application, the gear adjustment module comprises:

[0042] a second receiving submodule, configured to receive the control instruction sent by the voice recognition module;

[0043] a determining submodule, configured to determine a target working gear and adjust the target working gear.

[0044] According to some embodiments of the present application, a control method of a rechargeable hair dryer comprises:

[0045] The voice recognition module receives user voice data and processes the data, recognizes the processed user voice data, and determines a control instruction, which comprises a temperature control instruction and a wind force control instruction.

[0046] The gear adjusting module determines a target working gear according to the control instruction and adjusts the target working gear.

[0047] The present application receives user voice data, processes the user voice data, and effectively obtains clear voice data of the user in a noisy environment with multiple sound sources, thereby ensuring the accuracy of the obtained control instruction. Meanwhile, the voice is used for control and adjustment, which facilitates the convenience of the user for adjustment and makes the adjustment more intelligent. The gear adjusting module determines a target working gear according to the control instruction and adjusts the target working gear, which facilitates the response mode based on the instruction, quickly adjusts the gear, improves the control and adjustment efficiency of the rechargeable hair dryer, and thereby improves the satisfaction of the user.

[0048] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application will be realized and attained by the structure particularly pointed out in the written description and claims hereof.

[0049] The technical solutions of the present application will be further described in detail below with the aid of the accompanying drawings and embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0050] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, illustrate the present application, and explain the present application together with the embodiments of the present application, and do not constitute a limitation on the present application. In the drawings:

[0051] Figure 1 is a block diagram of a rechargeable hair dryer according to the present application;

[0052] Figure 2 is a block diagram of a voice recognition module according to an embodiment of the present application;

[0053] Figure 3 is a block diagram of a processing submodule according to an embodiment of the present application;

[0054] Figure 4 is a flow chart of a control method of a rechargeable hair dryer according to the present application. DETAILED DESCRIPTION

[0055] The preferred embodiments of the present application will be described in detail hereinafter with reference to the accompanying drawings, in which the preferred embodiments of the present application described herein are presented by way of illustration and explanation, and are not intended to limit the present application.

[0056] Embodiment 1:

[0057] The embodiment of the present application provides a rechargeable hair dryer, as shown in the accompanying drawings, comprising: Figure 1

[0058] a voice recognition module, configured to receive user voice data, perform data processing on the user voice data, recognize the user voice data after the data processing, and determine a control instruction; the control instruction comprises a temperature control instruction and a wind power control instruction;

[0059] a gear adjustment module, configured to receive the control instruction sent by the voice recognition module, determine a target working gear, and adjust the target working gear.

[0060] In the embodiment, the user voice data is voice data for adjusting the temperature or the wind power when the user uses the rechargeable hair dryer.

[0061] In the embodiment, the target working gear comprises a function type to be adjusted and a corresponding gear.

[0062] The working principle and beneficial effects of the above technical solution are as follows: by receiving the user voice data and performing data processing on the user voice data, clear voice data of the user can be effectively obtained in a noisy environment or an environment with multiple sound sources, so that the accuracy of the obtained control instruction is ensured. Meanwhile, the voice is used for control and adjustment, so that the convenience of the user for adjustment is improved, and the adjustment is more intelligent. The gear adjustment module determines the target working gear according to the control instruction and adjusts the target working gear, so that the response mode based on the instruction is facilitated, the gear adjustment is quickly performed, the control and adjustment efficiency of the rechargeable hair dryer is improved, and the satisfaction of the user is improved.

[0063] Embodiment 2:

[0064] Based on the embodiment 1, the rechargeable hair dryer further comprises:

[0065] a rechargeable battery pack;

[0066] a battery management module, configured to monitor the power storage and working state of the rechargeable battery pack, and manage the rechargeable battery pack based on the power storage and working state.

[0067] In the embodiment, the working state comprises working, standby and shutdown.​

[0068] The working principle and beneficial effects of the above technical solution are: by being equipped with a rechargeable battery pack, the hair dryer is freed from the use method of traditional wiring plug-in, so that the hair dryer can be more portable and flexible to operate. The battery management module monitors the storage and working state of the rechargeable battery pack, so that the user can know the state of the rechargeable battery pack in time, thereby charging or canceling the charging operation.

[0069] Embodiment 3:

[0070] On the basis of embodiment 1, the voice recognition module, as shown in Figure 2 , comprises:

[0071] The first receiving sub-module is used for receiving user voice data.

[0072] The processing sub-module is used for processing the user voice data to obtain the to-be-recognized user voice data.

[0073] The recognition sub-module is used for performing voice recognition on the to-be-recognized user voice data to determine the control instruction.

[0074] In this embodiment, the control instruction comprises a temperature control instruction and a wind force control instruction.

[0075] The working principle and beneficial effects of the above technical solution are: by receiving user voice data and processing the user voice data, clear to-be-recognized user voice data belonging to the user which can be subjected to voice recognition is obtained, so as to improve the accuracy of voice recognition on the to-be-recognized user voice data and ensure the accuracy of the control instruction.

[0076] Embodiment 4:

[0077] On the basis of embodiment 3, the processing sub-module, as shown in Figure 3 , comprises:

[0078] The noise reduction processing unit is used for reducing the noise signal in the user voice data to obtain first intermediate voice data.

[0079] The voice separation unit is used for separating the voice data of the user from the first intermediate voice data as second intermediate voice data.

[0080] The signal enhancement unit is used for performing high-frequency signal enhancement processing on the second intermediate voice data to obtain the to-be-recognized user voice data.

[0081] In this embodiment, the first intermediate voice data is the data after noise reduction of the user voice data.

[0082] In this embodiment, the second intermediate speech data is speech data of a user in a multi-source scene separated from the data after noise reduction.

[0083] In this embodiment, the user speech data to be recognized is data subjected to high-frequency signal enhancement processing of the second intermediate speech data.

[0084] The working principle and beneficial effects of the above technical solution are as follows: by performing noise reduction, separation and signal enhancement on the user speech data, high-quality speech data of the user in a noise scene or a multi-source scene is effectively obtained, which facilitates speech recognition, reduces the difficulty of speech recognition, and improves the recognition efficiency.

[0085] Embodiment 5:

[0086] On the basis of embodiment 3, the recognition sub-module comprises:

[0087] The feature extraction unit is configured to input the user speech data to be recognized into a pre-trained speech feature extraction model to obtain feature vector information of the user speech data to be recognized as the to-be-matched feature vector information.

[0088] The feature matching unit is configured to match the feature vector information with keyword speech feature vectors in a preset local speech library to obtain a matching result.

[0089] The semantic acquisition unit is configured to obtain a first user speech semantic based on the matching result and a semantic corresponding relationship corresponding to the matched keyword speech feature vector. The first user speech semantic comprises a function description and an adjustment metric description.

[0090] The judgment unit is configured to perform validity judgment on the first user speech semantic, take the first user speech semantic that passes the validity judgment as a target user speech semantic, and determine a control instruction according to the target user speech semantic.

[0091] In this embodiment, the validity judgment comprises judging the integrity and rationality of the function description and the adjustment metric description corresponding thereto. For example, when the first user speech semantic is "temperature, adjust to x degrees", firstly, it is judged that the corresponding relationship between "temperature" and "degrees" is complete, and secondly, it is judged whether "x degrees" is within the temperature adjustment range of the cordless hair dryer. If yes, the adjustment metric of "x degrees" is also reasonable, so that the first user speech semantic passes the validity judgment and can be taken as the target user speech semantic.

[0092] The working principle and beneficial effects of the above technical solutions are as follows: the feature extraction is performed on the to-be-recognized user voice data, so as to facilitate the matching of the obtained to-be-matched feature vector information with the keyword voice feature vectors in the local voice library and improve the matching speed. The first user voice semantics are obtained based on the obtained matching results and the semantic corresponding relationship corresponding to the matched keyword voice feature vectors, so as to facilitate the quick obtaining of the semantics of the user voice data. The effectiveness of the first user voice semantics is judged, so as to ensure the logic and control rationality of the first user voice semantics, thereby generating an effective control instruction and improving the satisfaction of the user using the rechargeable hair dryer.

[0093] Embodiment 6:

[0094] On the basis of embodiment 5, the judging unit comprises:

[0095] The first judging subunit is configured to judge the function description in the first user voice semantics and determine the function that needs to be adjusted.

[0096] The second judging subunit is configured to judge the adjustment metric description in the first user voice semantics after the first judging subunit determines the function that needs to be adjusted, and determine the target value.

[0097] The third judging subunit is configured to:

[0098] determine the adjustment range corresponding to the function that needs to be adjusted.

[0099] When it is determined that the target value is within the adjustment range, it is determined that the first user voice semantics that pass the effectiveness judgment are taken as the target user voice semantics through the effectiveness judgment, and the control instruction is determined according to the target user voice semantics.

[0100] In this embodiment, the function description comprises temperature and wind power.

[0101] In this embodiment, the adjustment metric description is the metric data for measuring the temperature corresponding to the temperature and the metric data for measuring the wind power corresponding to the wind power.

[0102] In this embodiment, the target value is represented as the target size of the temperature or the wind power.

[0103] The working principle and beneficial effects of the above technical solutions are as follows: the first judging subunit determines the function that needs to be adjusted, the second judging subunit determines the target value, and finally the third judging subunit determines the effectiveness of the function that needs to be adjusted and the target value, so as to obtain a reasonable and effective target user voice semantics, thereby ensuring the effectiveness of the generated control instruction.

[0104] Embodiment 7:

[0105] On the basis of embodiment 4, the noise reduction processing unit comprises:

[0106] The voice screening processing subunit is configured to remove pure silence data segments in the user voice data to obtain noise user voice data.

[0107] The voice data processing subunit is configured to determine noise frames and human voice frames in the noise user voice data.

[0108] The model training subunit is configured to train a target noise reduction neural network model based on training model samples.

[0109] The user voice noise reduction subunit is configured to obtain user noise reduction voice data based on the human voice frames and the target noise reduction neural network model, and take the user noise reduction voice data as the first intermediate voice data.

[0110] In this embodiment, pure silence data segments in the user voice data are removed to obtain noise user voice data, specifically including:

[0111] The user voice data is segmented and divided based on a voice fluctuation period of the user voice data to obtain a plurality of user voice data segments.

[0112] Each user voice data segment is layered to obtain a plurality of user voice data layers corresponding to each user voice data segment.

[0113] Voice amplitude data of the plurality of user voice data layers corresponding to each user voice data segment is obtained, and it is determined whether each user voice data segment is a pure silence data segment based on the voice amplitude data, and user voice data segments that are pure silence data segments are determined.

[0114] The positions of the pure silence data segments in the user voice data are determined, the pure silence data segments in the user voice data are removed, and noise user voice data is obtained.

[0115] In this embodiment, it is determined whether each user voice data segment is a pure silence data segment based on the voice amplitude data, that is, a user voice data segment with voice amplitude data of 0 is determined to be a pure silence data segment.

[0116] In this embodiment, noise frames and human voice frames in the noise user voice data are determined, specifically including:

[0117] The noise user voice data is divided into frames to obtain a plurality of user voice noise blocks.

[0118] The root mean square value of each user voice noise block included in the plurality of user voice noise blocks is calculated, and the maximum root mean square value is obtained.

[0119] Based on the maximum root mean square value, the root mean square value of each user voice noise block is normalized to obtain a corresponding judgment value.

[0120] The judgment value of each frame of user voice noise block is compared with the preset noise threshold value in sequence, and the user voice noise block with a judgment value less than the preset noise threshold value is determined as a noise frame;

[0121] The user voice noise block with a judgment value greater than or equal to the preset noise threshold value is determined as a human voice frame;

[0122] In this embodiment, based on the maximum root mean square value, the root mean square value of each frame of user voice noise block is normalized to obtain the corresponding judgment value, specifically: the root mean square value of each frame of user voice noise block is multiplied by a preset normalization parameter, and then divided by the maximum root mean square value, to obtain the root mean square value of each frame of user voice noise block after normalization, and the root mean square value is taken as the judgment value.

[0123] In this embodiment, the target noise reduction neural network model is trained based on the training model sample, specifically including:

[0124] The preset noisy speech data and the corresponding clean speech data are taken as the training model sample;

[0125] The noisy speech data in the training model sample is input into the noise reduction neural network model to obtain the first noise reduction speech data;

[0126] Based on the Fourier transform principle, the first noise reduction speech real part and the first noise reduction speech imaginary part corresponding to the first noise reduction speech data are obtained, and the clean speech real part and the clean speech imaginary part corresponding to the clean speech data are obtained;

[0127] The first noise reduction speech data and the clean speech data are loss calculated based on the si-snr loss function to obtain the first loss parameter;

[0128] The first noise reduction speech real part, the first noise reduction speech imaginary part, the clean speech real part and the clean speech imaginary part are calculated by the mean square loss to obtain the second loss parameter;

[0129] The weight parameter input by the model training personnel is obtained, the first loss parameter and the second loss parameter are weighted and added to obtain the weighted sum loss parameter, and the parameters of the noise reduction neural network model are corrected based on the weighted sum loss parameter, and the iteration is continued until the noise reduction neural network model converges to obtain the target noise reduction neural network model;

[0130] In the embodiment, the first noise-reduced speech real part, the first noise-reduced speech imaginary part, the pure speech real part, and the pure speech imaginary part are subjected to mean square loss calculation to obtain a second loss parameter. Specifically, based on the mean square error (MSE) loss calculation principle, the first noise-reduced speech real part and the pure speech real part and the first noise-reduced speech imaginary part and the pure speech imaginary part are taken as a group of calculation parameters, respectively, to calculate the overall loss to obtain the second loss parameter.

[0131] In the embodiment, based on the vocal frame and the target noise reduction neural network model, user noise-reduced speech data is obtained, and the user noise-reduced speech data is taken as first intermediate speech data, specifically including:

[0132] Based on the time sequence relationship of the user speech data, the plurality of vocal frames are sequentially input to the target noise reduction neural network model to obtain a plurality of noise-reduced vocal frames.

[0133] Based on the plurality of noise-reduced vocal frames, the user noise-reduced speech data is obtained, and the user noise-reduced speech data is taken as the first intermediate speech data.

[0134] The working principle and beneficial effects of the above technical solution are as follows: by screening out the noise user speech data in the user speech data, it is convenient to exclude the useless data in the user speech data, so as to save the speech noise reduction resources and improve the noise reduction efficiency. By determining the noise frame and the vocal frame in the noise user speech data, it is convenient to input the vocal frame to the trained target noise reduction neural network model, so as to quickly obtain the noise-reduced noise-reduced vocal frame, which not only can improve the noise reduction efficiency of the speech, but also can improve the noise reduction quality, and is convenient for improving the efficiency of subsequent identification of the user speech data.

[0135] Embodiment 8:

[0136] On the basis of embodiment 4, the speech separation unit includes:

[0137] The training subunit is configured to train the speech segmentation model based on the sample original speech data and the corresponding sample final speech data.

[0138] The separation subunit is configured to:

[0139] Based on the first intermediate speech data and the trained speech segmentation model, a segmented speech embedding feature sequence set of the first intermediate speech data is obtained.

[0140] Based on the segmented speech embedding feature sequence set, the first intermediate speech data is subjected to speech clustering to obtain the speech data of the user in the first intermediate speech data, and the speech data of the user is taken as second intermediate speech data.

[0141] In this embodiment, based on the sample original speech data and the corresponding sample final speech data, the speech segmentation model is trained, specifically including:

[0142] Obtain sample original speech data and corresponding sample final speech data;

[0143] Extract feature information of each speech frame in the sample original speech data, and generate an original speech feature map based on the feature information;

[0144] Construct a speech segmentation model and initialize the model parameters;

[0145] Input the original speech feature map into the speech segmentation model to obtain a plurality of segmentation feature maps of different segmentation dimensions;

[0146] Convolve the segmentation feature maps of different segmentation dimensions respectively to calculate a plurality of vector data corresponding thereto;

[0147] Based on the information of a plurality of elements included in the plurality of vector data, sample segmentation prediction data of the sample final speech data is obtained;

[0148] Obtain the labeled data of the sample final speech data in the original speech feature map;

[0149] Calculate the error between the sample segmentation prediction data and the labeled data to obtain an error result, and based on the error result, adjust the parameters of the speech segmentation model, and iteratively until the error result is less than a preset error threshold. The speech segmentation model with an error result less than the preset error threshold is used as the trained speech segmentation model;

[0150] In this embodiment, the speech segmentation model includes a basic segmentation model and a plurality of convolution layers corresponding to different convolution kernels.

[0151] In this embodiment, the plurality of vector data, each element in each vector data corresponds to a group of segmentation points, and the prediction information of the sample final original speech data is the value of each element.

[0152] In this embodiment, the labeled data includes labeled type data and corresponding labeled position data.

[0153] In this embodiment, the sample segmentation prediction data includes type prediction data and corresponding position prediction data.

[0154] In this embodiment, based on the first intermediate speech data and the trained speech segmentation model, a segmentation speech embedding feature sequence set of the first intermediate speech data is obtained, specifically including:

[0155] input the first intermediate speech data into the trained speech segmentation model to obtain a plurality of segmentation feature maps of different segmentation dimensions, and take the segmentation feature map conforming to the preset target dimension as a target segmentation feature map;

[0156] based on the target segmentation feature map, obtain a corresponding segmented speech feature information sequence set;

[0157] input the segmented speech feature information sequence set into a preset speech feature model to obtain a segmented speech embedding feature sequence set;

[0158] In this embodiment, based on the segmented speech embedding feature sequence set, the first intermediate speech data is subjected to speech clustering to obtain the speech data of the user in the first intermediate speech data, and the speech data is taken as the second intermediate speech data, which specifically comprises:

[0159] input the segmented speech embedding feature sequence set into a user speech clustering model previously constructed based on a neural network to obtain at least one sound source label prediction sequence set corresponding to the segmented speech embedding feature sequence set;

[0160] obtain a standard speech embedding feature corresponding to standard recognized speech data previously input by the user;

[0161] input the standard speech embedding feature into the user speech clustering model to obtain a standard sound source label;

[0162] match the sound source label prediction sequence set and the standard sound source label to obtain a sound source label prediction sub-sequence set;

[0163] obtain a plurality of speech segments of the first intermediate speech data corresponding to the sound source label prediction sub-sequence set;

[0164] based on the plurality of speech segments, sequentially restore and dereverberate the first intermediate speech data to obtain the speech data of the user in the first intermediate speech data, and take the speech data as the second intermediate speech data.

[0165] In this embodiment, the sound source label prediction sequence set is an ordered set in which the predicted identity labels of the sound sources are sequentially arranged.

[0166] In this embodiment, the standard sound source label is the identity label of the user of the cordless hair dryer.

[0167] In this embodiment, the sound source label prediction sub-sequence set is an ordered set in the sound source label prediction sequence set that is consistent with the standard sound source label.

[0168] The working principle and beneficial effects of the technical solution are as follows: the speech segmentation model is trained to obtain the segmented speech embedding feature sequence set from the first intermediate speech data, so as to improve the accuracy of obtaining the segmented speech embedding feature sequence set, the speech clustering is performed on the first intermediate speech data, the speech data of the user can be separated in the environment where multiple people speak, the accuracy of obtaining the speech data of the user is improved, the efficiency of the response of the rechargeable hair dryer to the user speech is improved, and the user experience is improved.

[0169] Embodiment 9:

[0170] On the basis of embodiment 1, the gear adjusting module comprises:

[0171] The second receiving sub-module is configured to receive the control instruction sent by the speech recognition module.

[0172] The determining sub-module is configured to determine the target working gear and adjust it.

[0173] The working principle and beneficial effects of the technical solution are as follows: the speech segmentation model is trained to obtain the segmented speech embedding feature sequence set from the first intermediate speech data, so as to improve the accuracy of obtaining the segmented speech embedding feature sequence set, the speech clustering is performed on the first intermediate speech data, the speech data of the user can be separated in the environment where multiple people speak, the accuracy of obtaining the speech data of the user is improved, the efficiency of the response of the rechargeable hair dryer to the user speech is improved, and the user experience is improved.

[0174] Embodiment 10:

[0175] According to some embodiments of the present application, a control method of a rechargeable hair dryer comprises the following steps as shown in Figure 4 The speech recognition module receives the user speech data and performs data processing, identifies the user speech data after data processing, and determines the control instruction; the control instruction comprises a temperature control instruction and a wind power control instruction.

[0176] The gear adjusting module determines the target working gear according to the control instruction and adjusts it.

[0177] In this embodiment, the user speech data is the speech data for adjusting the temperature or the wind power when the user uses the rechargeable hair dryer.

[0178] In this embodiment, the target working gear comprises the function type to be adjusted and the corresponding gear.

[0179] In this embodiment, the target working gear comprises the function type to be adjusted and the corresponding gear.

[0180] The working principle and beneficial effects of the above technical solution are as follows: by receiving user voice data and performing data processing on the user voice data, clear voice data of the user can be effectively obtained in a noisy and multi-sound source use environment, thereby ensuring the accuracy of the obtained control instruction. Meanwhile, the use of voice for control adjustment facilitates the convenience of user adjustment, making the adjustment more intelligent. The gear adjustment module determines the target working gear according to the control instruction and adjusts it, which facilitates the response mode based on the instruction, quickly adjusts the gear, improves the control and adjustment efficiency of the rechargeable hair dryer, and thereby improves the user's satisfaction.

[0181] Embodiment 11:

[0182] On the basis of embodiment 1, the method for determining the target working gear by the determining sub-module comprises:

[0183] When the control instruction is a temperature control instruction, the difference between the target value and the corresponding current value of the temperature control instruction is obtained;

[0184] The working power of the temperature control working element is determined based on the difference, and the target working gear is determined according to the working power.

[0185] The working principle and beneficial effects of the above technical solution are as follows: by obtaining the difference between the target value and the corresponding current value of the temperature control instruction, the target working gear is determined according to the working power of the temperature control working element, which facilitates the accuracy of determining the target working gear and ensures the safe and reliable operation of the rechargeable hair dryer.

[0186] Embodiment 12:

[0187] On the basis of embodiment 1, the rechargeable hair dryer further comprises:

[0188] The display module is used to display the temperature data and wind power data of the current gear in operation.

[0189] The working principle and beneficial effects of the above technical solution are as follows: by displaying the temperature data and wind power data, the user can intuitively understand the temperature data and wind power data of the current operation, and when the user has adjustment needs, the user can further determine the adjustment metric, thereby improving the user's experience.

[0190] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these modifications and variations.

Claims

1. A rechargeable hair dryer, characterized in that, include: The voice recognition module is used to receive user voice data, process the data, recognize the processed user voice data, and determine control commands; the control commands include temperature control commands and wind power control commands. The gear adjustment module is used to receive control commands sent by the voice recognition module, determine the target working gear, and adjust it accordingly. The speech recognition module includes: The first receiving submodule is used to receive user voice data; The processing submodule is used to process user voice data to obtain user voice data to be recognized; The recognition submodule is used to perform speech recognition on the user's speech data to be recognized and determine control commands. The processing submodule includes: A noise reduction processing unit is used to reduce noise signals in user voice data to obtain first intermediate voice data; The speech separation unit is used to separate the user's speech data from the first intermediate speech data and use it as the second intermediate speech data; The signal enhancement unit is used to perform high-frequency signal enhancement processing on the second intermediate speech data to obtain the user speech data to be recognized. The noise reduction processing unit includes: The voice filtering and processing subunit is used to remove clean and silent data segments from the user's voice data to obtain noisy user voice data. The speech data processing subunit is used to determine the noise frames and human voice frames in the speech data of noisy users. The model training subunit is used to train the target denoising neural network model based on the training model samples; The user speech noise reduction subunit is used to obtain user noise reduction speech data based on human voice frames and target noise reduction neural network models, and to use the user noise reduction speech data as the first intermediate speech data. The voice data processing subunit removes clean, silent segments from the user's voice data to obtain noisy user voice data, specifically including: Based on the voice fluctuation cycle of user voice data, the user voice data is segmented to obtain several user voice data segments. Each user voice data segment is divided into layers to obtain several user voice data layers corresponding to each user voice data segment; Obtain the voice amplitude data of several user voice data layers corresponding to each user voice data segment, and determine whether each user voice data segment is a pure silent data segment based on the voice amplitude data, and identify the user voice data segment that is a pure silent data segment. The location of the clean silence data segment in the user's voice data is determined, and the clean silence data segment in the user's voice data is removed to obtain the noisy user voice data. Identifying noise frames and human voice frames in noisy user speech data specifically includes: The noisy user speech data is divided into frames to obtain multi-frame user speech noise blocks; Calculate the root mean square value of each frame of user speech noise block included in the multi-frame user speech noise block, and obtain the largest root mean square value among them. Based on the maximum root mean square value, the root mean square value of each frame of user speech noise block is normalized to obtain the corresponding judgment value. The judgment value of each frame of user speech noise block is compared with the preset noise threshold in turn, and the user speech noise block with the judgment value less than the preset noise threshold is determined as a noise frame. User speech noise blocks with a judgment value greater than or equal to a preset noise threshold are identified as human voice frames.

2. A rechargeable hair dryer as described in claim 1, characterized in that, The rechargeable hair dryer also includes: Rechargeable battery pack; The battery management module is used to monitor the energy storage and operating status of the rechargeable battery pack, and to manage the rechargeable battery pack based on the energy storage and operating status.

3. A rechargeable hair dryer as described in claim 1, characterized in that, The identification submodule includes: The feature extraction unit is used to input the user's speech data to be identified into a pre-trained speech feature extraction model to obtain the feature vector information of the user's speech data to be identified, which is used as the feature vector information to be matched. The feature matching unit is used to match the feature vector information with the keyword speech feature vectors in the preset local speech database to obtain the matching result; The semantic acquisition unit is used to obtain the first user's speech semantics based on the matching results and the semantic correspondence between the matched keyword speech feature vectors; the first user's speech semantics includes a functional description and a regulation metric description. The judgment unit is used to judge the validity of the first user's speech semantics, take the first user's speech semantics that has passed the validity judgment as the target user's speech semantics, and determine the control command based on the target user's speech semantics.

4. A rechargeable hair dryer as described in claim 3, characterized in that, The judgment unit includes: The first judgment subunit is used to judge the functional description in the first user's speech semantics and determine the function that needs to be adjusted; The second judgment subunit is used to judge the adjustment metric description in the first user's speech semantics and determine the target value after the first judgment subunit determines the function that needs to be adjusted. The third decision subunit is used for: Determine the adjustment range corresponding to the function that needs to be adjusted; When the target value is determined to be within the adjustment range, the validity judgment is determined, and the first user's voice semantics that passed the validity judgment is taken as the target user's voice semantics. The control command is determined based on the target user's voice semantics.

5. A rechargeable hair dryer as described in claim 1, characterized in that, The speech separation unit includes: The training subunit is used to train the speech segmentation model based on the original speech data of the samples and the corresponding final speech data of the samples. Separate subunits, used for: Based on the first intermediate speech data and the trained speech segmentation model, the segmented speech embedding feature sequence set of the first intermediate speech data is obtained; Based on the segmented speech embedding feature sequence set, speech clustering is performed on the first intermediate speech data to obtain the user's speech data in the first intermediate speech data, which is then used as the second intermediate speech data.

6. A rechargeable hair dryer as described in claim 1, characterized in that, The gear adjustment module includes: The second receiving submodule is used to receive control commands sent by the speech recognition module; The determination submodule is used to determine the target working level and make adjustments.

7. A control method for a rechargeable hair dryer, characterized in that, include: The system receives user voice data based on a voice recognition module, processes the data, recognizes the processed user voice data, and determines control commands; the control commands include temperature control commands and wind power control commands. The gear adjustment module determines the target working gear and adjusts it according to the control command; The speech recognition module includes: The first receiving submodule is used to receive user voice data; The processing submodule is used to process user voice data to obtain user voice data to be recognized; The recognition submodule is used to perform speech recognition on the user's speech data to be recognized and determine control commands. The processing submodule includes: A noise reduction processing unit is used to reduce noise signals in user voice data to obtain first intermediate voice data; The speech separation unit is used to separate the user's speech data from the first intermediate speech data and use it as the second intermediate speech data; The signal enhancement unit is used to perform high-frequency signal enhancement processing on the second intermediate speech data to obtain the user speech data to be recognized. The noise reduction processing unit includes: The voice filtering and processing subunit is used to remove clean and silent data segments from the user's voice data to obtain noisy user voice data. The speech data processing subunit is used to determine the noise frames and human voice frames in the speech data of noisy users. The model training subunit is used to train the target denoising neural network model based on the training model samples; The user speech noise reduction subunit is used to obtain user noise reduction speech data based on human voice frames and target noise reduction neural network models, and to use the user noise reduction speech data as the first intermediate speech data. The clean, silent segments of user voice data are removed to obtain noisy user voice data, specifically including: Based on the voice fluctuation cycle of user voice data, the user voice data is segmented to obtain several user voice data segments. Each user voice data segment is divided into layers to obtain several user voice data layers corresponding to each user voice data segment; Obtain the voice amplitude data of several user voice data layers corresponding to each user voice data segment, and determine whether each user voice data segment is a pure silent data segment based on the voice amplitude data, and identify the user voice data segment that is a pure silent data segment. The location of the clean silence data segment in the user's voice data is determined, and the clean silence data segment in the user's voice data is removed to obtain the noisy user voice data. Identifying noise frames and human voice frames in noisy user speech data specifically includes: The noisy user speech data is divided into frames to obtain multi-frame user speech noise blocks; Calculate the root mean square value of each frame of user speech noise block included in the multi-frame user speech noise block, and obtain the largest root mean square value among them. Based on the maximum root mean square value, the root mean square value of each frame of user speech noise block is normalized to obtain the corresponding judgment value. The judgment value of each frame of user speech noise block is compared with the preset noise threshold in turn, and the user speech noise block with the judgment value less than the preset noise threshold is determined as a noise frame. User speech noise blocks with a judgment value greater than or equal to a preset noise threshold are identified as human voice frames.

Images