CN109688036B

CN109688036B - Control method and device of intelligent household appliance, intelligent household appliance and storage medium

Info

Publication number: CN109688036B
Application number: CN201910126388.2A
Authority: CN
Inventors: 宋夏
Original assignee: Guangzhou Shiyuan Electronics Thecnology Co Ltd
Current assignee: Guangzhou Shiyuan Electronics Thecnology Co Ltd
Priority date: 2019-02-20
Filing date: 2019-02-20
Publication date: 2021-12-03
Anticipated expiration: 2039-02-20
Also published as: CN109688036A

Abstract

The invention discloses a control method and device of an intelligent household appliance, the intelligent household appliance and a storage medium. The method comprises the following steps: receiving voice information; acquiring voice command words, wherein the voice command words are obtained by inputting voice information and a voice recognition algorithm into a voice recognition model, and the voice recognition model is determined according to the working scene of the intelligent household appliance; and controlling the intelligent household appliance according to the voice command words. When the intelligent household appliance is in a working scene with small calculation amount, the voice recognition model with small calculation amount is used, so that the hardware cost, the CPU calculation resource and the CPU power consumption of the intelligent household appliance are reduced; when the intelligent household appliance is in a working scene with a large calculation amount, the voice recognition model with the large calculation amount is used in a self-adaptive mode, so that CPU calculation resources are reasonably utilized, and the system response speed and the overall power consumption of the working process of the intelligent household appliance are guaranteed to be good.

Description

Control method and device of intelligent household appliance, intelligent household appliance and storage medium

Technical Field

The embodiment of the invention relates to the technology of intelligent household appliances, in particular to a control method and device of an intelligent household appliance, the intelligent household appliance and a storage medium.

Background

With the rapid development of science and technology, more and more intelligent household appliances are added with a voice recognition function. In order to ensure the voice recognition rate, a voice recognition algorithm with a high recognition rate generally needs to be adopted, and correspondingly, the complexity of a voice recognition model corresponding to the voice recognition algorithm with the high recognition rate is higher, so that when the voice recognition algorithm is operated, a larger CPU (central processing unit) operation resource is consumed, and further the problems of slow response and high power consumption of a system of the intelligent household appliance are caused.

Disclosure of Invention

The invention provides a control method and device of an intelligent household appliance, the intelligent household appliance and a storage medium, and aims to solve the problems of slow system response and high power consumption during voice control of the intelligent household appliance.

In a first aspect, an embodiment of the present invention provides a method for controlling an intelligent home appliance, including:

receiving voice information;

acquiring voice command words, wherein the voice command words are obtained by inputting voice recognition algorithms into the voice information and the voice recognition model, and the voice recognition model is determined according to the working scene of the intelligent household appliance;

and controlling the intelligent household appliance according to the voice command word.

Further, the searching precision of the voice recognition algorithm is determined according to the working performance of the intelligent household appliance.

Further, the working performance of the intelligent household appliance at least comprises one of the CPU utilization rate and the CPU power consumption of the intelligent household appliance.

Further, the determining, according to the working performance of the intelligent household appliance, the search accuracy of the voice recognition algorithm includes:

acquiring the current process running time of the intelligent household appliance;

and calculating a pruning coefficient of the voice recognition algorithm, wherein the pruning coefficient is determined according to the CPU utilization rate of the intelligent household appliance, the CPU utilization rate is the proportion of the current process running time to the total CPU running time, and the search precision is in inverse proportion to the pruning coefficient.

Further, before the obtaining the voice command word, the method further includes:

training a voice recognition model, wherein training input parameters of the voice recognition model comprise simulation working scene information and simulation voice command words of the intelligent household appliance.

Further, the speech recognition model includes an acoustic model and a language model.

Further, the working scene of the intelligent household appliance is determined according to at least one of the load working mode, the background running program and the current time of the intelligent household appliance.

In a second aspect, an embodiment of the present invention further provides a control device for an intelligent home appliance, including:

the receiving module is used for receiving voice information;

the acquisition module is used for acquiring voice command words, wherein the voice command words are obtained by inputting voice recognition algorithms into the voice information and the voice recognition model, and the voice recognition model is determined according to the working scene of the intelligent household appliance;

and the control module is used for controlling the intelligent household appliance according to the voice command words.

Further, the control device for the intelligent household appliance further comprises:

and the determining module is used for determining the searching precision of the voice recognition algorithm according to the working performance of the intelligent household appliance.

Further, the determining module comprises:

the acquisition unit is used for acquiring the current process running time of the intelligent household appliance;

and the calculation unit is used for calculating the pruning coefficient of the voice recognition algorithm, the pruning coefficient is determined according to the CPU utilization rate of the intelligent household appliance, the CPU utilization rate is the proportion of the current process running time to the total CPU running time, and the search precision is in inverse proportion to the pruning coefficient.

Further, the control device of the intelligent household appliance further comprises:

and the training module is used for training a voice recognition model before the voice command words are obtained, and the training input parameters of the voice recognition model comprise the simulated work scene information and the simulated voice command words of the intelligent household appliance.

In a third aspect, an embodiment of the present invention further provides an intelligent home appliance, including: a display screen, a memory, and one or more processors;

the display screen is used for displaying the state information of the intelligent household appliance;

the memory for storing one or more programs;

when the one or more programs are executed by the one or more processors, the one or more processors implement the method for controlling an intelligent appliance according to the first aspect.

In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the method for controlling an intelligent appliance according to the first aspect.

The voice recognition method comprises the steps of receiving voice information input by a user, determining a proper voice recognition model according to a working scene of the intelligent household appliance, inputting the voice information and the voice recognition model into a voice recognition algorithm to obtain a voice command word, and controlling the intelligent household appliance according to the voice command word. When the intelligent household appliance is in a working scene with small operand, the voice recognition model with small operand is used, so that the hardware cost, CPU (central processing unit) operation resources and CPU power consumption of the intelligent household appliance are reduced; when the intelligent household appliance is in a working scene with a large calculation amount, the voice recognition model with the large calculation amount is used in a self-adaptive mode, so that CPU calculation resources are reasonably utilized, and the system response speed and the overall power consumption of the working process of the intelligent household appliance are guaranteed to be good.

Drawings

Fig. 1 is a flowchart of a control method for an intelligent household appliance according to an embodiment of the present invention;

fig. 2 is a flowchart of a control method for an intelligent home appliance according to a second embodiment of the present invention;

fig. 3 is a flowchart of a control method for an intelligent home appliance according to a third embodiment of the present invention;

fig. 4 is a schematic structural diagram of a plurality of intelligent home appliances according to a third embodiment of the present invention;

fig. 5 is a schematic diagram of a television program pause display provided in the third embodiment of the present invention;

fig. 6 is a schematic structural diagram of a control device of an intelligent home appliance according to a fourth embodiment of the present invention;

fig. 7 is a schematic structural diagram of an intelligent home appliance according to a fifth embodiment of the present invention.

Detailed Description

The present invention will be described in further detail with reference to the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting of the invention. It should be further noted that, for the convenience of description, only some of the structures related to the present invention are shown in the drawings, not all of the structures.

Example one

Fig. 1 is a flowchart of a method for controlling an intelligent home appliance according to an embodiment of the present invention, where the method for controlling an intelligent home appliance according to the embodiment may be executed by an intelligent home appliance, and the intelligent home appliance includes two or more physical entities or is a single physical entity. For example, intelligent appliances include, but are not limited to, intelligent range hoods, intelligent air conditioners, intelligent televisions, intelligent water heaters, and the like.

As shown in fig. 1, the method for controlling an intelligent home appliance specifically includes the following steps:

and S110, receiving voice information.

The voice information refers to a voice operation instruction sent to the intelligent household appliance by a user. The language of the voice message is not limited, for example, mandarin, english or various dialects; the sentence sequencing, the sentence type and the like of the voice information are not limited, for example, the voice information may be "turn off the fan", or the voice information may be "turn the air volume up", and the like.

In an embodiment, a user sends a voice message to a smart appliance to wake up the smart appliance. Such as: when a user cooks, the current air volume of the intelligent range hood cannot meet the air volume requirement, and then the user can speak out voice information such as 'turn large air volume' and the like to the intelligent range hood, and turn large air volume of the intelligent range hood through the voice information.

And S120, acquiring the voice command word.

The voice command words are obtained by inputting voice information and a voice recognition model into a voice recognition algorithm, and the voice recognition model is determined according to the working scene of the intelligent household appliance.

The voice information input by the user is generally spoken instruction information, acoustic features of the voice information are recognized through a voice recognition model to recognize the voice information input by the user, and the voice information and the voice recognition model are input to a predetermined voice recognition algorithm to obtain a standard command word corresponding to the voice information, namely a voice command word. For example, assuming that the voice information input by the user is "turn down the air volume of the range hood to the first gear", the voice command word received by the intelligent range hood is "turn down the air volume to the first gear".

Specifically, after the intelligent household appliance receives voice information input by a user, the intelligent household appliance can extract acoustic features in the voice information, input the acoustic features into a predetermined voice recognition model, recognize the acoustic features through the voice recognition model to recognize the voice information input by the user, and input the voice information and the voice recognition model into a predetermined voice recognition algorithm to obtain a corresponding voice command word.

Wherein the speech recognition model comprises an acoustic model and a language model. Where an acoustic model is a knowledge representation of differences in acoustics, linguistics, variables of the environment, speaker gender, accents, etc., and a language model is a knowledge representation of a set of word sequences. Generally speaking, an acoustic model is trained by lstm + ctc to obtain mapping from acoustic features to phonemes; the language model uses SRILM tool to train LM to obtain 3-gram and4-gram, which is the mapping between words and sentences. For a speech recognition system, the output values of the acoustic model are acoustic features calculated for each frame to map each phoneme in the speech information; and the speech model is a mapping of words and sentences in the speech information.

The working scene of the intelligent household appliance can be understood as the current use scene of the intelligent household appliance. Particularly, when the intelligent household appliances are different terminal devices, the working scenes included in the intelligent household appliances are different. For example, when the intelligent household appliance is an intelligent range hood, the working scenes of the intelligent household appliance can include a standby scene, a cooking scene, an audio and video playing scene and the like; when the intelligent household appliance is an intelligent air conditioner, the working scenes of the intelligent household appliance can comprise a standby scene, a timing working scene, a refrigeration scene, a heating scene and the like; when the intelligent household appliance is an intelligent television, the working scene of the intelligent household appliance can include a standby scene, an audio and video playing scene and the like, and the working scene of each intelligent household appliance is not limited in the embodiment. The audio and video playing scene can be an independent music playing scene or a scene in which audio and video are used simultaneously.

In an embodiment, the working scene of the intelligent household appliance is determined according to at least one of the load working mode, the background running program and the current time of the intelligent household appliance. The load working mode may be understood as a working mode in which each load in the intelligent household appliance is located, for example, the load working mode is a working operation mode and a standby mode; the background running program can be understood as a related program of communication running in the intelligent household appliance; the current time is the current Beijing time. In general, the intelligent range hood can be considered to be in a cooking scene when the current time is 6-8 am, or 11-12 pm or 5-7 pm. Exemplarily, assuming that the intelligent household appliance is an intelligent range hood, the load of the intelligent range hood can include a fan, a lighting lamp, a display screen and the like, the fan is in a working operation mode, the current time is 7 am, and the working scene of the intelligent range hood can be judged to be a cooking scene. And when the display screen of the intelligent range hood is in a working operation mode, the background operation program is in a communication interaction stage, the fan and the illuminating lamp are in a closed state, and only the loudspeaker in the range hood is in an open state, the working scene of the intelligent range hood is judged to be an audio and video playing scene. Of course, when the intelligent household appliance is a different terminal device, different conditions are required to be adopted to determine the working scene, and in the embodiment, the determination of the working scene is described only by taking the intelligent range hood as an example, but not limited thereto.

It should be noted that a preset database is provided in the intelligent household appliance, and different pre-trained speech recognition models are stored in the preset database. In an embodiment, when the intelligent home appliance is in different working scenes, the corresponding voice recognition models are different. Wherein each speech recognition model corresponds to a designated work scenario. Exemplarily, assuming that the intelligent household appliance is an intelligent range hood, the working scenes of the intelligent range hood may include a standby scene, a cooking scene, an audio and video playing scene, and the like. Of course, under different working scenarios, the high-probability used command word list and the low-probability used command word list in the corresponding speech recognition model are completely different. For example, when the working scene of the intelligent range hood is a cooking scene, a voice recognition model corresponding to the cooking scene is selected from a preset database, and in the cooking scene, because a fan needs to be turned on to process oil smoke or a lighting lamp is turned on to illuminate the environment where the intelligent range hood is located, command words related to control of the fan and lighting can be included in a command word list with high probability, and command words related to control such as switching to the next video and pausing playing can be included in a command word list with low probability.

And S130, controlling the intelligent household appliance according to the voice command word.

Specifically, after the intelligent household appliance obtains the voice command word, corresponding command operation is executed according to the voice command word, so that the working scene of the intelligent household appliance is adaptively adjusted to meet the use requirement of a user. For example, when the intelligent range hood is in a cooking scene, that is, the fan is in an open state, at this time, a user can send voice information such as "close the fan", the voice information and a voice recognition model corresponding to the cooking scene are input into a predetermined voice recognition algorithm to obtain a voice command word such as "close the fan", and after the intelligent range hood receives the voice command word of "close the fan", the operating mode of the fan is switched from open to closed according to the voice command word to realize voice control over the intelligent range hood. It should be noted that converting "turning off the fan" into "turning off the fan" is a process of converting voice into text, that is, converting voice information input by a user into a corresponding voice command word.

According to the technical scheme of the embodiment, the voice information input by the user is received, the appropriate voice recognition model is determined according to the working scene of the intelligent household appliance, then the voice information and the voice recognition model are input into the voice recognition algorithm to obtain the voice command word, and the intelligent household appliance is controlled according to the voice command word. When the intelligent household appliance is in a working scene with small operand, the voice recognition model with small operand is used, so that the hardware cost, CPU (central processing unit) operation resources and CPU power consumption of the intelligent household appliance are reduced; when the intelligent household appliance is in a working scene with a large calculation amount, the voice recognition model with the large calculation amount is used in a self-adaptive mode, so that CPU calculation resources are reasonably utilized, and the system response speed and the overall power consumption of the working process of the intelligent household appliance are guaranteed to be good.

Example two

Fig. 2 is a flowchart of a control method for an intelligent home appliance according to a second embodiment of the present invention. The present embodiment is a method for controlling an intelligent home appliance, which is based on the above-described embodiments. As shown in fig. 2, the method for controlling an intelligent home appliance of the present embodiment specifically includes the following steps:

and S210, training a voice recognition model.

It should be noted that when the user inputs the voice information, the voice information received by the intelligent home appliance may be affected by the surrounding environment and include noise, or the user inputs a certain accent when the voice information, or the voice information sent by the user may have a situation that it is unclear to tell a word, and the like. Therefore, in order to improve the recognition rate of the speech recognition model, the speech recognition model may be trained before receiving the speech information.

The training input parameters of the voice recognition model comprise simulated work scene information and simulated voice command words of the intelligent household appliance. In an embodiment, the simulated work scene information is information for simulating that the intelligent household appliance is in different work scenes. Specifically, when the intelligent household appliance is in different working scenes, the surrounding environment where the intelligent household appliance is located is also different, for example, when the intelligent range hood is in a cooking scene for simulation, because the fan is in an on state, noise exists in the surrounding environment where the intelligent range hood is located, in order to enable the intelligent range hood to obtain voice information input by a user, the voice information can be more accurately recognized, when an acoustic model is trained, noise generated by operation of the fan needs to be considered, and noise adding training processing is performed on a simulated voice command word.

In the embodiment, the simulated voice command word is a voice command word with different scores for the intelligent household appliance in different working scenes, and can be understood as a voice command word in the simulated working scene. In order to improve the voice recognition rate of the intelligent household appliance, when the simulation work scene information and the simulation voice command words are trained, the simulation voice command words can be set according to different simulation work scenes. Specifically, according to different working scenes where the intelligent household appliance is located, a corresponding high-probability used voice command word list and a corresponding low-probability used voice command word list in the working scene are determined, and then the simulated working scene information and the simulated voice command words are trained according to the selected voice command word list. Each voice command word list corresponds to one simulated work scene, and meanwhile, the voice command word list comprises a plurality of simulated voice command words.

When the intelligent household appliance is in different working scenes and the acoustic model and the language model are trained, different simulated voice command words can be trained, and the voice command words in the voice command word list used at a high probability and the voice command word list used at a low probability are different.

Taking the cooking scene of the intelligent range hood as an example, the training speech recognition model is specifically explained. Aiming at a cooking scene, because the scores of voice command words related to the control of the emphasis fans such as the fan and the illuminating lamp are higher, the voice command word list used at high probability can comprise analog voice command words related to the control of the emphasis fans such as the fan turning-off, the fan turning-on, the air volume turning-up, the air volume turning-down and the first/second/third gear air volume, and the voice command word list used at low probability can comprise analog voice command words related to the control of the emphasis music such as the volume turning-up, the next song, the previous song, the pause and the play. Meanwhile, when the acoustic model is subjected to simulation training, in order to enable the simulation working scene information of the simulation training to be closer to the working scene in the actual operation process, the noise of the fan operation needs to be considered, and the noise-adding training needs to be carried out while the simulation voice command word is input. Certainly, when the language model is trained, the scores of the simulated voice command words related to the fan control of the emphasis are improved in the language model, the scores of the language models corresponding to other simulated voice command words are reduced, and then the search time of the voice recognition algorithm and the CPU operation resources are reduced.

Specifically, it is assumed that the score of the simulated voice command word ranges from 0 to 100 points, where 0 indicates that the obtained voice command word is completely mismatched with the simulated voice command word of the intelligent appliance, and 100 points indicate that the obtained voice command word is completely matched with the simulated voice command word of the intelligent appliance, where a preset threshold of the simulated voice command word in the voice command word list with high probability of use is set to 80 points and is recorded as a first preset threshold, and a preset threshold of the simulated voice command word in the voice command word list with low probability of use is set to 60 points and is recorded as a second preset threshold. Illustratively, when the intelligent range hood is in a cooking scene, the scores of the simulated voice command words related to the control of the emphasis fans, such as 'turn-off fan', 'turn-on fan', 'turn-up air quantity', 'turn-down air quantity' and 'first/second/third air quantity', are set to be more than 80 minutes, and the scores of the simulated voice command words related to the control of the emphasis music, such as 'turn-up volume, next song, previous song, pause, play', are set to be less than 60 minutes, namely, the scores of the language models corresponding to the simulated voice command words related to the control of the emphasis music are reduced, so that the search time of the voice recognition models is reduced, and further the waste of CPU (Central processing Unit) operation resources is reduced.

It should be noted that, when the acoustic model is simulated and trained for different working scenarios, the setting may be performed through the score of the acoustic model, and may also be embodied in the setting of the search path. Illustratively, when the intelligent range hood is in a cooking scene, special processing can be carried out on simulated voice command words in the cooking scene. For example, a simulated voice command word similar to "air volume turning up" such as "component turning up", "air volume turning up", etc. is also corresponding to the voice command word "air volume turning up", thereby increasing the voice recognition rate of the voice command word. Meanwhile, special processing is not needed to be carried out on the analog voice command words related to the music control, such as 'the previous song' and 'the switching to the next video', so that the CPU computing resources are effectively utilized.

Of course, when the voice recognition model is trained, in order to ensure the accuracy of the voice recognition rate in the simulation working environment, only one intelligent household appliance is provided in the same environment. However, in an actual application scenario, there is not only one intelligent home appliance. For example, at home of a certain user, terminal devices such as an intelligent range hood, an intelligent air conditioner, an intelligent television and an intelligent water heater are generally installed, and at this moment, the situation that the same voice command word may be applicable to a plurality of terminal devices occurs, so that the false triggering rate of the intelligent household appliance can be caused.

In order to ensure the voice recognition rate of the intelligent household appliance and increase the diversity, a plurality of simulated voice command words can be selected to respectively train a plurality of voice recognition models, and each voice recognition model is independent. Meanwhile, in order to reduce the false triggering rate of the intelligent household appliance, when the voice recognition model is trained, the simulation voice command words can be correspondingly set. For example, when the intelligent range hood and the intelligent television are in an audio and video playing scene, command words related to music control such as volume, program, next/previous command word and the like exist in a command word list used at a high probability, and when a voice recognition model of the intelligent range hood in the audio and video playing scene is trained, simulated voice command words can be set as words such as 'volume of the range hood is reduced' and 'the range hood is switched to the next' and the like; correspondingly, when the intelligent television is trained on the voice recognition model in the audio and video playing scene, the analog voice command word can be set to be similar words such as 'volume of the television is turned down' and 'the television is switched to the next channel', namely, the name and the operation command of the intelligent household appliance are set in the analog voice command word.

Of course, in the embodiment, only the name and the operation command of the intelligent household appliance are exemplarily set in the simulated voice command word, and a specific voice command word may also be used to represent one intelligent household appliance, so as to distinguish different intelligent household appliances, for example, the intelligent water heater is named as "love", so that when the intelligent water heater is trained, love and a corresponding operation command need to be set in the simulated voice command word.

And S220, receiving voice information.

And S230, acquiring voice command words.

The searching precision of the voice recognition algorithm is determined according to the working performance of the intelligent household appliance. In an embodiment, the operating performance of the intelligent appliance includes at least one of a CPU usage rate and a CPU power consumption of the intelligent appliance.

The CPU utilization rate is a CPU operation resource occupied by the current running process and represents the running process condition of the intelligent household appliance at a certain time point. Generally speaking, the CPU utilization rate is the situation that all running processes occupy the CPU computing resource, and is not the situation that a single running process occupies the CPU computing resource. Specifically, the higher the CPU utilization rate is, the more processes the intelligent appliance runs at this time point is, and vice versa, the less processes the intelligent appliance runs at this time point is. Of course, the CPU utilization rate is not strictly proportional to the number of currently running processes, for example, there are two processes currently running, but the CPU consumption of each process is 30%; of course, it is also possible that currently 10 processes are running, but the CPU consumption of each process is only 1%. Specifically, when the utilization rate of the CPU is increased, it indicates that the speech recognition algorithm consumes a large amount of CPU operation resources, and at this time, the search accuracy of the speech recognition algorithm should be appropriately adjusted to reduce the utilization rate of the CPU, so that the phenomena of seizure, serious heating, and the like of the smart home appliance are improved; on the contrary, when the CPU utilization rate is reduced, the voice recognition algorithm consumes less CPU operation resources, and the search precision of the voice recognition algorithm can be properly improved.

Correspondingly, the power consumption of the CPU is the product of the current value flowing through the processor core of the intelligent household appliance and the core voltage value on the processor, and is also a judgment condition for the cooling system of the intelligent household appliance. Generally, if the CPU power consumption is low, it means that the CPU generates less heat, and indirectly means that the CPU utilization is also low; conversely, if the CPU power consumption is high, it means that the CPU generates a large amount of heat, and indirectly means that the CPU utilization is also high. In general, the power consumption of the CPU increases because the operating frequency of the CPU itself in the intelligent home appliance is increased, for example, when the operating frequency of the CPU itself in the intelligent home appliance is increased from 600MHz to 1GHz, the power consumption of the CPU also increases. Meanwhile, the working frequency is in direct proportion to the utilization rate of the CPU, and it can be understood that the current working frequency of the intelligent household appliance cannot meet the requirement due to the high utilization rate of the CPU, and the working frequency of the intelligent household appliance needs to be increased.

Specifically, the searching accuracy of the speech recognition algorithm is determined according to the working performance of the intelligent household appliance, and the method specifically comprises the following steps of S10-S20:

and S10, acquiring the current process running time of the intelligent household appliance.

The current process running time is the time of a CPU occupied by the current process in the intelligent household appliance. Generally, when the time-sharing multitasking operating system uses the CPU in a time-sharing mode, the CPU is not occupied by the process B when the process A occupies the CPU. In the embodiment, when the intelligent household appliance is in an open state, each load, background running program and foreground running program in the intelligent household appliance have corresponding processes, and the running time of the current process in the intelligent household appliance can be obtained by calculating the running time of the current process in the intelligent household appliance according to the calculation mode of the current process running time in the prior art.

And S20, calculating the pruning coefficient of the voice recognition algorithm.

The pruning coefficient is determined according to the CPU utilization rate of the intelligent household appliance, the CPU utilization rate is the proportion of the current process running time to the total CPU running time, and the search precision is inversely proportional to the pruning coefficient.

In an embodiment, CPU usage is the ratio of current process run time to the total CPU run time. Exemplarily, it is assumed that an a process and a B process run in the intelligent home appliance, where the a process occupies 10ms, the B process occupies 30ms, and the B process is idle for 60 ms; then the process A takes 10ms, the process B takes 30ms, and the process B is idle for 60ms, if all the processes are in a period of time, the CPU occupancy rate in the period of time is 40%. Wherein, the CPU occupancy rate is the CPU utilization rate.

The pruning coefficient is a parameter used for pruning the decision tree pruning algorithm in the voice recognition algorithm. Specifically, the search accuracy of the speech recognition algorithm is determined by the working performance of the intelligent household appliance on one hand and is related to the complexity of the speech recognition algorithm on the other hand. When the pruning coefficient is higher, the complexity of the speech recognition algorithm is indicated to be lower, and conversely, when the pruning coefficient is lower, the complexity of the speech recognition algorithm is indicated to be higher. The pruning coefficient is determined according to the CPU utilization rate of the intelligent household appliance, specifically, the relation between the pruning coefficient and the CPU utilization rate is related to a voice recognition algorithm and a voice recognition model adopted by the intelligent household appliance at present, and the operation of the voice recognition algorithm is reduced according to the relation along with the increase of the pruning coefficient. In the embodiment, the higher the pruning coefficient is, the more easily the branch path having a low score is pruned when the optimal path search is performed in the voice recognition algorithm, so that the CPU computation amount is reduced, but the search accuracy is also reduced. Illustratively, the relation between the pruning coefficient and the CPU utilization is B ═ a × C + B, where B is the pruning coefficient, C is the current CPU utilization of the smart appliance, and a and B are empirical coefficients. Wherein, A and B are obtained by testing through a large amount of experiments in the development stage. It can be seen that the pruning coefficient and the CPU utilization rate are in a linear relationship, and when the pruning coefficient B increases, the empirical coefficients a and B are adjusted so that the CPU utilization rate C decreases accordingly.

It should be noted that, when a certain voice command word is searched in the voice recognition algorithm, there are several search paths, and in order to reduce CPU computation resources when the intelligent home appliance runs the voice recognition algorithm, branch paths with a low probability in the voice recognition algorithm need to be cut out. Specifically, there are several layers in the decision tree of the speech recognition algorithm, and there are several leaf nodes in each layer, so that in the process of cutting branch paths, the branch paths need to be cut according to a preset pruning coefficient, and branch paths with a probability lower than the pruning coefficient are cut, so as to reduce the use of CPU computing resources. For example, assuming that the pruning coefficient is 0.9, three branch paths in the speech recognition algorithm are A, B and C, wherein the probabilities of A, B and C are 90%, 85% and 80%, respectively, when the branch paths are pruned, the B path and C path are pruned, so that the optimal search path in the speech recognition algorithm is identified, and the search time is reduced.

It should be noted that, in the embodiment, the operational relationship between the pruning coefficient and the CPU utilization is listed only by way of example, and the pruning coefficient and the CPU utilization are in a linear relationship, but are not limited thereto.

And S240, controlling the intelligent household appliance according to the voice command word.

According to the technical scheme of the embodiment, on the basis of the embodiment, the CPU utilization rate is obtained by obtaining the current process running time of the intelligent household appliance, the pruning coefficient of the voice recognition algorithm is determined according to the CPU utilization rate, and the branch path with the low probability in the voice recognition algorithm is cut out, so that the search precision of the voice recognition algorithm is dynamically adjusted according to the condition of the current CPU operation resource, the CPU operation resource is reasonably utilized, and the CPU utilization rate of the intelligent household appliance is further improved.

EXAMPLE III

Fig. 3 is a flowchart of a control method for an intelligent home appliance according to a third embodiment of the present invention. In addition to the above embodiments, the present embodiment specifically describes a control method of an intelligent home appliance as a preferred embodiment. The control method specifically comprises the following steps:

and S310, judging the use scene of the current equipment.

The equipment is terminal equipment, such as intelligent range hood, intelligent air conditioner, intelligent television and other intelligent household appliances. In the embodiment, the use scene is a work scene of the current intelligent household appliance. Specifically, when determining the working scene of the current intelligent household appliance, various information of the intelligent household appliance, such as the running application programs of the foreground and the background of the current intelligent household appliance, the running conditions of each load in the current intelligent household appliance, the current time and the like, needs to be collected first, and the working scene is determined according to the information. Illustratively, when intelligent household electrical appliances are intelligent lampblack absorber, can gather the operational aspect of loads such as fan, light among the intelligent household electrical appliances, when the fan is in the operating mode in service, can judge that this intelligent lampblack absorber is in the culinary art scene. Or, when the current time is 6 pm or half, it can be determined that the intelligent range hood is in a cooking scene.

Of course, in order to ensure the accuracy of determining the working scene, a plurality of information of the intelligent household appliance may be collected, for example, the working scene of the intelligent household appliance may be determined according to the running condition of the load, the current time and the foreground running program. Exemplarily, the lighting lamp of the intelligent range hood is in an on state, the current time is 5 o' clock and half afternoon, but the fan is in an off state, a video playing interface is displayed on the display screen, and meanwhile, the loudspeaker also sends out audio information, so that the intelligent range hood can be judged to be in an audio and video playing scene.

S320, querying a database and selecting a proper algorithm model.

Wherein, a pre-trained speech recognition model is stored in the database. In the embodiment, the algorithm model is a voice recognition model used when the voice recognition algorithm is operated, and specifically, after the working scene of the intelligent household appliance is determined, a preset database can be queried according to the relevant information of the current working scene of the intelligent household appliance, and a proper acoustic model and a proper voice model are selected. The preset database is confirmed in the equipment development stage, and the working scene is firstly divided according to the type of the equipment, for example, when the equipment is an intelligent range hood, the working scene can be divided into a standby scene, a cooking scene, an audio and video playing scene and the like. And determining a command word list with high probability of use and a command word list with low probability of use in each different working scene. And training a corresponding voice recognition model aiming at the selected command word list. For example, aiming at a cooking scene, the acoustic model mainly trains command words related to fan control, noise of fan operation is considered, noise adding training is carried out, scores of the command words related to fan control are improved in the language model, and meanwhile related recognition rejection paths of voice command words related to fan control are reduced, so that the CPU (Central processing Unit) operation amount is reduced. The recognition rejection path may be understood as a path of "recognition rejection" added to the language model to prevent a near-speech word from causing misrecognition of the speech command word. For example, for the voice command word of "turn on the fan", a rejection path similar to "turn on the fan" may be set to prevent the smart appliance from recognizing the voice command word as "turn on the fan" when the user sends out the voice message of "turn on the fan" and causing a malfunction of the smart appliance. Particularly, in a cooking scene, because a user needs to control and operate the fan, the use probability of the voice command word of turning on the fan is higher in the scene, and if relevant recognition rejection paths corresponding to the relevant voice command words controlled by the fan are reduced, the misoperation probability of the intelligent household appliance is reduced, so that the search paths of the voice recognition algorithm in the process of searching the voice command words corresponding to the voice information are reduced, the search time is shortened, and the CPU (central processing unit) operation resources are saved.

Illustratively, the intelligent range hood with the voice recognition function has two types of preset simulated voice command words of range hood fan control and music control, for example, the simulated voice command words of range hood fan control include: the method comprises the following steps of turning off a fan, adjusting the air volume to be large, adjusting the air volume to be third gear air volume and the like; the music-controlled simulated voice command words include: volume up, next song, etc. When the situation that a user opens the intelligent range hood is monitored, the voice recognition algorithm selects an acoustic model which is trained aiming at a fan control command and is subjected to noise adding training, and selects a language model which emphasizes fan control related command words; and when the user is monitored to open the music player interface, the voice recognition algorithm selects the acoustic model and the language model which are optimized with emphasis on the music control command.

And S330, calculating the current CPU utilization rate.

Specifically, the current CPU utilization rate of the intelligent household appliance is calculated so as to calculate the pruning coefficient of the voice recognition algorithm. The method for calculating the CPU utilization rate may be various, for example, the CPU utilization rate is calculated according to a ratio of idle running time to total CPU running time, and the CPU utilization rate may also be calculated according to a ratio of current process running time to total CPU running time.

And S340, selecting a proper pruning coefficient according to the current CPU utilization rate.

In the embodiment, according to the CPU utilization rate of the current intelligent household appliance, the pruning coefficient when the optimal branch path is searched in the voice recognition algorithm is determined. The higher the pruning coefficient is, the easier the best branch path search is, the branch path with the lower probability is cut, so that the CPU calculation amount is reduced, and correspondingly, the search precision of the voice recognition algorithm is also reduced. Specifically, when it is monitored that the CPU consumption of the current intelligent household appliance is relatively high, which indicates that the CPU utilization of the current intelligent household appliance is relatively high, the pruning coefficient of the voice recognition algorithm needs to be increased appropriately to reduce the search accuracy and further reduce the CPU consumption; on the contrary, when the CPU consumption is lower, which indicates that the CPU utilization rate of the current intelligent household appliance is lower, the pruning coefficient of the voice recognition algorithm needs to be properly reduced so as to improve the search accuracy of the voice recognition algorithm.

And S350, operating a voice recognition algorithm to recognize the voice command of the user.

Specifically, after a voice recognition model and a pruning coefficient of a voice recognition algorithm are determined according to a working scene of the intelligent household appliance, the voice recognition model and voice information sent by a user are input into the voice recognition algorithm to operate the voice recognition algorithm, the voice information sent by the user is recognized, a voice command word is obtained, and the voice control is carried out on the intelligent household appliance.

According to the technical scheme of the embodiment, a proper voice recognition model is selected according to the use scene of the intelligent household appliance, and the consumption of CPU (Central processing Unit) operation resources is reduced as much as possible on the premise of ensuring the voice recognition rate; meanwhile, the search precision of the voice recognition algorithm is dynamically adjusted according to the CPU utilization rate of the current intelligent household appliance, and the utilization rate of the CPU is improved.

Fig. 4 is a schematic structural diagram of a plurality of intelligent home appliances according to a third embodiment of the present invention. As shown in fig. 4, a method for controlling an intelligent household appliance will be specifically described with four terminal devices, i.e., an intelligent range hood 301, an intelligent air conditioner 302, an intelligent television 303, and an intelligent water heater 304, being disposed in the home of a certain user.

Suppose that the user 300 sends out a voice message of "pause a television program" to perform voice control on the smart television 303, but if the working environment of the smart range hood 301 is also in an audio/video playing scene at this time, and a command word list used at a high probability in the voice recognition model also contains a simulated voice command word similar to "pause", then the corresponding smart home appliance needs to be selected according to the device name in the voice command word at this time to perform voice control on the smart home appliance. Specifically, after the user 300 sends the voice message, the four terminal devices in the current environment all receive the voice message, and since the intelligent range hood 301 and the intelligent television 303 are both in the audio/video playing scene, when receiving the voice message, the intelligent television 303 inputs the voice message and a voice recognition model determined in advance according to the audio/video playing scene into a voice recognition algorithm to obtain a corresponding voice command word, and then pauses the program of the intelligent television according to the voice command word. Fig. 5 is a schematic display diagram of television program pause according to a third embodiment of the present invention, and as shown in fig. 5, an interface of a pause key 3032 is displayed on a display screen 3031 of the smart television 303 to prompt a user that the television program has been paused, so that the user can visually observe an effect after the smart television 303 executes a voice command word on the display screen; when the intelligent range hood 301 receives the voice message, since the voice message does not include the name of the "range hood," the intelligent range hood 301 does not obtain a corresponding voice command word according to the voice message, that is, the intelligent range hood 301 does not respond.

Of course, in the embodiment, the control method of the intelligent household appliance is described by taking "pause the television program" and the intelligent range hood 301, the intelligent air conditioner 302, the intelligent television 303 and the intelligent water heater 304 as examples. In the actual operation process, the user may set the operation according to the actual situation, which is not limited.

Example four

Fig. 6 is a schematic structural diagram of a control device of an intelligent home appliance according to a fourth embodiment of the present invention. The controlling means of intelligence household electrical appliances that this embodiment provided can integrate in intelligent household electrical appliances, and this intelligence household electrical appliances can be that two or more physical entities constitute, also can be that a physical entity constitutes, and this intelligence household electrical appliances can be intelligent household electrical appliances, for example, intelligent lampblack absorber, intelligent air conditioner, intelligent TV, intelligent water heater etc.. Referring to fig. 6, the control device for an intelligent home appliance provided in this embodiment specifically includes: a receiving module 410, an obtaining module 420, and a control module 430.

The receiving module 410 is configured to receive voice information;

the acquiring module 420 is configured to acquire a voice command word, where the voice command word is obtained by inputting a voice recognition algorithm into a voice information and voice recognition model, and the voice recognition model is determined according to a working scene of the intelligent household appliance;

and the control module 430 is used for controlling the intelligent household appliance according to the voice command word.

On the basis of the above embodiment, the control device for an intelligent home appliance further includes:

On the basis of the above embodiment, the working performance of the intelligent appliance at least includes one of a CPU usage rate and a CPU power consumption of the intelligent appliance.

On the basis of the above embodiment, the determining module further includes:

and the computing unit is used for computing the pruning coefficient of the voice recognition algorithm, the pruning coefficient is determined according to the CPU utilization rate of the intelligent household appliance, the CPU utilization rate is the proportion of the current process running time to the total CPU running time, and the search precision is in inverse proportion to the pruning coefficient.

and the training module is used for training the voice recognition model before the voice command word is obtained, and the training input parameters of the voice recognition model comprise the simulated work scene information and the simulated voice command word of the intelligent household appliance.

On the basis of the above-described embodiment, the speech recognition model includes an acoustic model and a language model.

On the basis of the embodiment, the working scene of the intelligent household appliance is determined according to at least one of the load working mode, the background running program and the current time of the intelligent household appliance.

The control device of the intelligent household appliance can execute the control method of the intelligent household appliance provided by any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

EXAMPLE five

Fig. 7 is a schematic structural diagram of an intelligent home appliance according to a fifth embodiment of the present invention. This intelligent household electrical appliance includes: a processor 510, a memory 520, an input device 530, an output device 540, and a display screen 550. The number of the processors 510 in the intelligent appliance may be one or more, and one processor 510 is taken as an example in fig. 7. The number of the memories 520 in the intelligent appliance may be one or more, and one memory 520 is illustrated in fig. 7. The processor 510, the memory 520, the input device 530, the output device 540, and the display screen 550 of the intelligent appliance may be connected by a bus or other means, and fig. 7 illustrates the connection by the bus as an example. In the embodiment, the intelligent household appliance can be an intelligent range hood, an intelligent air conditioner, an intelligent television, an intelligent water heater and the like.

The memory 520 is a computer-readable storage medium for storing software programs, computer-executable programs, and modules, such as program instructions/modules corresponding to the intelligent appliance according to any embodiment of the present invention (for example, the receiving module 410, the obtaining module 420, and the control module 430 in the control device of the intelligent appliance). The memory 520 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application program required for at least one function; the storage data area may store data created according to use of the device, and the like. Further, the memory 520 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid state storage device. In some examples, memory 520 may further include memory located remotely from processor 510, which may be connected to devices through a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The input means 530 may be used to receive input numeric or character information and generate key signal inputs related to user settings and function control of the intelligent appliance, and may also be a camera for acquiring images and a sound pickup device for acquiring audio data. The output device 540 may include an audio device such as a speaker. It should be noted that the specific composition of the input device 530 and the output device 540 can be set according to actual situations.

The display screen 550 may be a touch-enabled display screen 550, which may be a capacitive screen, an electromagnetic screen, or an infrared screen. Generally, the display screen 550 is configured to display data according to an instruction of the processor 510, for example, to display status information of the smart appliance, for example, to display information of power amount and operation mode, and to receive a touch operation applied to the display screen 550, for example, when the smart appliance is a range hood, the touch operation may be an option of increasing an air volume, switching operation modes, and sending a corresponding signal to the processor 510 or other devices.

The processor 510 executes various functional applications and data processing of the device by executing software programs, instructions and modules stored in the memory 520, so as to implement the above-mentioned control method of the intelligent home appliance.

The intelligent household appliance provided by the above can be used for executing the control method of the intelligent household appliance provided by any of the above embodiments, and has corresponding functions and beneficial effects.

EXAMPLE six

An embodiment of the present invention further provides a storage medium containing computer-executable instructions, where the computer-executable instructions are executed by a computer processor to perform a method for controlling an intelligent appliance, and the method includes:

receiving voice information;

acquiring a voice command word, wherein the voice command word is obtained by inputting voice information and a voice recognition model into a voice recognition algorithm, and the voice recognition model is determined according to the working scene of the intelligent household appliance;

and controlling the intelligent household appliance according to the voice command words.

Of course, the storage medium containing the computer-executable instructions provided in the embodiments of the present invention is not limited to the above-described operations of the control method of the intelligent appliance, and may also perform related operations in the control method of the intelligent appliance provided in any embodiments of the present invention, and has corresponding functions and advantages.

From the above description of the embodiments, it is obvious for those skilled in the art that the present invention can be implemented by software and necessary general hardware, and certainly, can also be implemented by hardware, but the former is a better embodiment in many cases. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, and the computer software product may be stored in a computer-readable storage medium, such as a floppy disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a FLASH Memory (FLASH), a hard disk or an optical disk of a computer, and includes several instructions to enable a computer device (which may be a robot, a personal computer, a server, or a network device) to execute the method for testing the VBO display interface according to any embodiment of the present invention.

It should be noted that, in the control device of the intelligent household electrical appliance, each unit and each module included in the control device are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, specific names of the functional units are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present invention.

It should be understood that portions of the present invention may be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the various steps or methods may be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or combination of the following techniques, which are known in the art, may be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application specific integrated circuit having an appropriate combinational logic gate circuit, a Programmable Gate Array (PGA), a Field Programmable Gate Array (FPGA), or the like.

In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present invention and the technical principles employed. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the present invention has been described in greater detail by the above embodiments, the present invention is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A control method of an intelligent household appliance is characterized by comprising the following steps:

receiving voice information;

controlling the intelligent household appliance according to the voice command word;

the working scene of the intelligent household appliance is determined according to at least one of a background running program and the current time of the intelligent household appliance; the searching precision of the voice recognition algorithm is determined according to the working performance of the intelligent household appliance; the working performance of the intelligent household appliance at least comprises one of the CPU utilization rate and the CPU power consumption of the intelligent household appliance;

wherein, the step of determining the searching precision of the voice recognition algorithm according to the working performance of the intelligent household appliance comprises the following steps:

calculating a pruning coefficient of the voice recognition algorithm, wherein the pruning coefficient is determined according to the CPU utilization rate of the intelligent household appliance, the CPU utilization rate is the proportion of the current process running time to the total CPU running time, and the search precision is in inverse proportion to the pruning coefficient;

the pruning coefficient is determined according to the CPU utilization rate of the intelligent household appliance and comprises the following steps:

if the CPU utilization rate of the current intelligent household appliance is higher, the pruning coefficient of the voice recognition algorithm is increased; and if the CPU utilization rate of the current intelligent household appliance is low, the pruning coefficient of the voice recognition algorithm is reduced.

2. The method of claim 1, prior to said obtaining a voice command word, further comprising:

3. The method of claim 1, wherein the speech recognition model comprises an acoustic model and a language model.

4. The method of claim 1, wherein the working scenario of the smart appliance is further determined according to a load working mode of the smart appliance.

5. A control device of an intelligent household appliance, comprising:

the receiving module is used for receiving voice information;

the control module is used for controlling the intelligent household appliance according to the voice command words;

6. An intelligent appliance, comprising: a display screen, a memory, and one or more processors;

the memory for storing one or more programs;

when executed by the one or more processors, cause the one or more processors to implement the method of controlling an intelligent appliance according to any one of claims 1-4.

7. A computer-readable storage medium on which a computer program is stored, the program, when executed by a processor, implementing the method of controlling an intelligent appliance according to any one of claims 1-4.