Voice command recognition method, device, equipment and storage medium for intelligent elevator

Abstract

The present disclosure relates to a method, apparatus, device and storage medium for recognizing voice commands of an intelligent elevator, the method comprising: collecting audio emitted by a user in an elevator car; recognizing command words in the audio; determining whether a first audio segment before a target audio segment and / or a second audio segment after the target audio segment in the audio are emitted by the user who emitted the target audio segment, the target audio segment being an audio segment containing a command word in the audio; if not, determining that the command word is valid, and executing the corresponding elevator command according to the command word. The present disclosure takes into account that the elevator commands spoken by the user are usually concise and have fewer words. If, in the collected audio, the audio segment before the command word and / or the audio segment after the command word are spoken by the same user as the command word, it is considered that the user is chatting, and the intention of speaking the command word to take the elevator is not credible. Otherwise, the command word is considered valid, thereby reducing the probability of the intelligent elevator system being mistakenly awakened by voice commands.

Description

Technical Field

[0001] The present disclosure relates to the field of speech recognition technology, and in particular to a method, device, equipment, and storage medium for voice command recognition of an intelligent elevator. Background Art

[0002] When a smart elevator calls for help, the voice collection device collects the audio of the speaker in the car. The elevator call system then uses automatic speech recognition technology to perform voice recognition on the collected audio. When the recognition results contain elevator call commands such as "go to the first floor" or "close the door", the elevator call system responds accordingly to reach the designated floor or close the elevator door.

[0003] Since users inevitably talk in elevators, the collected audio often contains words with similar pronunciations to command words, or words that contain command words but no intention to take the elevator, causing the elevator call system to be falsely awakened. Summary of the Invention

[0004] In order to solve the above technical problems or at least partially solve the above technical problems, the present disclosure provides a voice command recognition method, device, equipment and storage medium for an intelligent elevator.

[0005] In a first aspect, the present disclosure provides a method for recognizing voice commands of an intelligent elevator, comprising:

[0006] Collect audio from users in the elevator car;

[0007] identifying a command word in the audio;

[0008] determining whether a first audio segment preceding a target audio segment and / or a second audio segment following the target audio segment in the audio is uttered by a user who uttered the target audio segment, wherein the target audio segment is the audio segment containing the command word in the audio;

[0009] If not, it is determined that the command word is valid, and the corresponding elevator instruction is executed according to the command word.

[0010] Optionally, before identifying the command word in the audio, the method further includes:

[0011] Perform noise reduction processing on the audio.

[0012] Optionally, the identifying the command word in the audio includes:

[0013] Converting the audio into corresponding text;

[0014] The text is semantically compared with a preset command word, and a word in the text that matches the preset command word is determined to be the command word.

[0015] Optionally, converting the audio into corresponding text includes:

[0016] The audio is input into an automatic speech recognition model to obtain the text output by the automatic speech recognition model.

[0017] Optionally, the text is marked with the pronunciation time of each word in the audio;

[0018] The determining whether the first audio segment preceding the target audio segment and / or the second audio segment following the target audio segment in the audio is sent by the user who sent the target audio segment includes:

[0019] Locating the starting time point and the ending time point of the target audio segment in the audio according to the pronunciation time corresponding to each character of the command word in the text;

[0020] Using the audio of a preset duration before the starting time point as the first audio segment;

[0021] Using the audio of a preset duration after the end time point as the second audio segment;

[0022] Comparing the first audio segment and the second audio segment with the target audio segment for voiceprint features, respectively, to obtain corresponding comparison results;

[0023] Determine whether the first audio segment and / or the second audio segment is uttered by the user who uttered the target audio segment based on the corresponding comparison results.

[0024] Optionally, the performing voiceprint feature comparison on the first audio segment and the second audio segment with the target audio segment respectively to obtain corresponding comparison results includes:

[0025] Inputting the first audio segment, the second audio segment, and the target audio into a voiceprint feature extraction model respectively to obtain a first voiceprint feature, a second voiceprint feature, and a third voiceprint feature respectively;

[0026] determining a first similarity between the first voiceprint feature and the third voiceprint feature, and a second similarity between the second voiceprint feature and the third voiceprint feature;

[0027] The first similarity and the second similarity are taken as a comparison result.

[0028] Optionally, determining whether the first audio segment and / or the second audio segment is sent by the user who sent the target audio segment according to the corresponding comparison result includes:

[0029] If the first similarity is greater than a preset threshold, determining that the first audio segment is sent by the user who sent the target audio segment;

[0030] If the second similarity is greater than a preset threshold, it is determined that the second audio segment is emitted by the user who emitted the target audio segment.

[0031] In a second aspect, the present disclosure provides a voice command recognition device for an intelligent elevator, comprising:

[0032] A collection module, used to collect audio emitted by users in the elevator car;

[0033] A recognition module, configured to recognize command words in the audio;

[0034] a determination module, configured to determine whether a first audio segment preceding a target audio segment and / or a second audio segment following the target audio segment in the audio is uttered by a user who uttered the target audio segment, wherein the target audio segment is the audio segment containing the command word in the audio;

[0035] The execution module is used to determine whether the command word is valid and execute the corresponding elevator instruction according to the command word.

[0036] Optionally, the device further includes a noise reduction module for performing noise reduction processing on the audio before recognizing the command word in the audio.

[0037] Optionally, when identifying a command word in the audio, the recognition module is specifically used to convert the audio into corresponding text; perform a semantic comparison between the text and a preset command word, and determine that the word in the text that matches the preset command word is the command word.

[0038] Optionally, when converting the audio into corresponding text, the recognition module is specifically used to input the audio into an automatic speech recognition model to obtain the text output by the automatic speech recognition model.

[0039] Optionally, the text is marked with the pronunciation time of each word in the audio; when the determination module determines whether the first audio segment before the target audio segment and / or the second audio segment after the target audio segment in the audio is issued by the user who issued the target audio segment, it is specifically used to locate the starting time point and the ending time point of the target audio segment in the audio according to the pronunciation time corresponding to each word of the command word in the text; use the audio of a preset length before the starting time point as the first audio segment; use the audio of a preset length after the end time point as the second audio segment; perform voiceprint feature comparison on the first audio segment and the second audio segment with the target audio segment respectively to obtain corresponding comparison results; and determine whether the first audio segment and / or the second audio segment is issued by the user who issued the target audio segment based on the corresponding comparison results.

[0040] Optionally, when the determination module compares the voiceprint features of the first audio segment and the second audio segment with the target audio segment respectively to obtain corresponding comparison results, the determination module is specifically used to input the first audio segment, the second audio segment and the target audio into a voiceprint feature extraction model respectively to obtain a first voiceprint feature, a second voiceprint feature and a third voiceprint feature respectively; determine a first similarity between the first voiceprint feature and the third voiceprint feature, and a second similarity between the second voiceprint feature and the third voiceprint feature; and use the first similarity and the second similarity as the comparison result.

[0041] Optionally, when the determination module determines whether the first audio segment and / or the second audio segment is issued by the user who issued the target audio segment based on the corresponding comparison results, it is specifically used to determine that the first audio segment is issued by the user who issued the target audio segment if the first similarity is greater than a preset threshold; if the second similarity is greater than the preset threshold, it is determined that the second audio segment is issued by the user who issued the target audio segment.

[0042] In a third aspect, the present disclosure provides an electronic device, comprising:

[0043] Memory;

[0044] processor; and

[0045] computer programs;

[0046] The computer program is stored in the memory and is configured to be executed by the processor to implement the method as described in the first aspect.

[0047] In a fourth aspect, the present disclosure provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method as described in the first aspect.

[0048] The technical solution provided by the present disclosure has the following advantages compared with the existing technology:

[0049] The present disclosure provides a method, apparatus, device and storage medium for recognizing voice commands for smart elevators. After collecting audio from a user in an elevator car, the method identifies command words in the audio, and then determines whether the first audio segment before the target audio segment and / or the second audio segment after the target audio segment corresponding to the command word in the audio are spoken by the user who spoke the target audio segment. If not, the target command word is determined to be valid, and the elevator command corresponding to the command word is executed. The principle is that, considering that the elevator voice commands spoken by the user are usually concise and have fewer words, if the audio segment before the command word and / or the audio segment after the command word in the collected audio are spoken by the same user as the command word, it is considered that the user is chatting and the intention of taking the elevator by speaking the command word is not credible. On the contrary, the intention of taking the elevator expressed by the command word is considered credible. In this way, the user's intention of taking the elevator can be determined more accurately, making the recognition of elevator voice commands more accurate, thereby reducing the probability of the smart elevator system being mistakenly awakened by voice commands. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0051] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0052] Figure 1 A flow chart of a method for recognizing voice commands in an intelligent elevator according to an embodiment of the present disclosure;

[0053] Figure 2 A schematic diagram of an application scenario provided by an embodiment of the present disclosure;

[0054] Figure 3 A schematic diagram of the structure of a voice command recognition device for an intelligent elevator provided in an embodiment of the present disclosure;

[0055] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0056] In order to more clearly understand the above-mentioned objectives, features and advantages of the present disclosure, the scheme of the present disclosure will be further described below. It should be noted that the embodiments of the present disclosure and the features therein can be combined with each other in the absence of conflict.

[0057] In the following description, many specific details are set forth to facilitate a full understanding of the present disclosure, but the present disclosure may also be implemented in other ways different from those described herein; it is obvious that the embodiments in the specification are only part of the embodiments of the present disclosure, rather than all of the embodiments.

[0058] Because conversations inevitably occur in elevators, the collected audio often contains phrases with similar pronunciations to command words, or phrases that contain command words but do not indicate an intention to board the elevator, causing the elevator call system to falsely wake up. To address these issues, the present disclosure provides a method for voice command recognition in an intelligent elevator. This method is described below with reference to specific examples.

[0059] Figure 1 This is a flow chart of the voice command recognition method for an intelligent elevator provided by an embodiment of the present disclosure. The method can be executed by a voice command recognition device for an intelligent elevator, which can be implemented in software and / or hardware, and can be configured in an electronic device, such as a server or terminal. In addition, the method can be applied to Figure 2 The application scenario shown includes an elevator car 201, a user 202, a user 203, a sound receiving device 204, and a central control computer 205. It is understandable that the voice command recognition method for an intelligent elevator provided in the embodiment of the present disclosure can also be applied in other scenarios.

[0060] The following combination Figure 2 The application scenario shown is Figure 1 The voice command recognition method of the intelligent elevator shown in the figure is introduced, for example, Figure 2 The central control computer 205 in the embodiment can execute the method. The method comprises the following steps:

[0061] S101. Collect audio signals emitted by users in an elevator car.

[0062] For example, the central control computer 205 continuously collects the audio emitted by the user 202 and the user 203 in the elevator car 201 through the audio receiving device 204. In this scenario, the user 202 and the user 203 are talking.

[0063] S102: Recognize command words in the audio.

[0064] The central control computer 205 uses automatic speech recognition technology (ASR) to recognize command words in the audio collected by the audio receiving device 204. For example, the audio of the conversation between user 202 and user 203 includes the question from user 202 to user 203, "Are you going to the sixth floor? I remember where you live." The central control computer 205 recognizes the command word <go to the sixth floor> in the audio.

[0065] S103: Determine whether a first audio segment preceding a target audio segment and / or a second audio segment following a target audio segment in the audio is sent by a user who sends the target audio segment, where the target audio segment is an audio segment containing a command word.

[0066] The central control computer 205 determines the audio segment in which the command word is issued in the collected audio as the target audio segment, and based on the position of the target audio segment in the collected audio, cuts off a segment of a preset length before the position as the first audio segment, and cuts off a segment of a preset length after the position as the second audio segment, wherein the first audio segment and the second audio segment do not overlap with the target audio segment.

[0067] Taking into account that the voice instructions issued by the user when taking the elevator are generally concise and contain fewer words, by determining whether the first audio segment and / or the second audio segment is issued by the user who issued the target audio segment, it is determined whether the user's intention to take the elevator expressed in the voice instruction of the command word is credible. If so, it is considered that the intention to take the elevator expressed by the user when speaking the command word is uncredible. If not, it is considered that the intention to take the elevator expressed by the user when speaking the command word is credible.

[0068] For example, the command word identified in S102 is <go to the sixth floor>, and the corresponding target audio segment is "go to the sixth floor". Based on the position of the target audio segment corresponding to the command word in the collected audio, the first audio segment intercepted by the central control computer 205 is "Are you ready?" and the second audio segment is "I'll remember". By performing voiceprint comparison between the first audio segment and the target audio segment, it is determined that the first audio segment and the target audio segment are both spoken by user 202. In this case, the intention of user 202 to take the elevator when speaking the command word <go to the sixth floor> is considered unreliable. Similarly, after performing voiceprint comparison between the second audio segment and the target audio segment and determining that the second audio segment and the target audio segment are both spoken by user 202, the intention of user 202 to take the elevator when speaking the command word <go to the sixth floor> can also be considered unreliable. In other words, as long as one of the first audio segment and the second audio segment is the same user who spoke the target audio segment, the intention of the command word to take the elevator is considered unreliable, thereby reducing the probability of the intelligent elevator system being falsely awakened. On the contrary, if the first audio segment and the second audio segment are not spoken by the same person as the target audio segment, the intention to take the elevator expressed by the command word is considered credible, and the next step is executed.

[0069] As described above, if not, execute S104.

[0070] S104: Determine whether the command word is valid, and execute the corresponding elevator instruction according to the command word.

[0071] If central control computer 205 determines that both the first and second audio segments are not spoken by the same user as the target audio segment, it deems the command word to represent the elevator intention credible. Central control computer 205 determines that the command word is valid and executes the corresponding elevator command according to the command word. For example, user 202 says "go to the sixth floor," and user 203 asks "what to go for." In this case, the first audio segment before the target audio segment "go to the sixth floor" is unvoiced, while the second audio segment after the target audio segment is "what to go for," which is spoken by user 203. Central control computer 205 determines that the first and second audio segments were not spoken by user 202, and therefore determines that the command word "go to the sixth floor" is valid. Central control computer 205 executes the elevator command to the sixth floor, moving elevator car 201 to the sixth floor.

[0072] The embodiment of the present disclosure collects audio of a user in an elevator car, identifies the command words in the audio, and then determines whether the first audio segment before the target audio segment and / or the second audio segment after the target audio segment corresponding to the command word in the audio are spoken by the user who spoke the target audio segment. If not, the target command word is determined to be valid, and the elevator instruction corresponding to the command word is executed. The principle is that, considering that the elevator voice instructions spoken by the user are usually concise and have fewer words, if the audio segment before the command word and / or the audio segment after the command word in the collected audio are spoken by the same user as the command word, it is considered that the user is chatting and the intention of taking the elevator by speaking the command word is not credible. On the contrary, the intention of taking the elevator expressed by the command word is considered credible. In this way, the user's intention of taking the elevator can be determined more accurately, and the recognition of elevator voice instructions is more accurate, thereby reducing the probability of the intelligent elevator system being mistakenly awakened by voice instructions.

[0073] Optionally, before recognizing the command words in the audio, the method further includes: performing noise reduction processing on the audio.

[0074] For example, before performing automatic speech recognition on the audio collected by the sound receiving device 204, the central control computer 205 first performs noise reduction processing on the audio to reduce the impact of noise on speech recognition. For example, a minimum mean square error noise reduction algorithm or a noise reduction model can be used to reduce the noise of the collected audio, or a linear filter can be used to reduce noise when the sound receiving device collects the audio.

[0075] The embodiments of the present disclosure reduce the impact of noise on speech recognition by performing noise reduction processing on the collected user audio.

[0076] Based on the above implementation, identifying command words in audio includes: converting the audio into corresponding text; performing semantic comparison between the text and preset command words, and determining that the words in the text that match the preset command words are command words.

[0077] When the central control computer 205 identifies command words in the collected audio, it first converts the voice content contained in the audio into corresponding text. It then performs a semantic comparison between the converted text and the preset command words to obtain a result that matches the words in the text with the preset command words, and uses the words in the matching result as the command words in the audio. For example, the vocabulary includes preset command words such as "go" and "X floor", where "X" is an integer. When the user 202 says "go up to the sixth floor", the central control computer 205 considers "up" to be semantically similar to "go" in the context of elevator instructions, and "sixth floor" is a specific floor. Therefore, it can be determined that "go up to the sixth floor" is a result of a semantic match with the preset command word, and thus "go up to the sixth floor" is recognized as a command word.

[0078] In this case, converting the audio into the corresponding text includes: inputting the audio into an automatic speech recognition model to obtain the text output by the automatic speech recognition model.

[0079] Automatic speech recognition technology is a technology that converts human speech into text, and it is often implemented through a model. A widely applicable automatic speech recognition model can be trained through a deep neural network. For example, the datatang_asr model is a trained open-source model, which can be directly used without the steps of obtaining a dataset and training the model. In addition, an automatic speech recognition model corresponding to the application scenario can also be trained. The central control computer 205 outputs the audio collected by the radio device 204 into the automatic speech recognition model, and uses the automatic speech recognition model to perform speech recognition to obtain the output text.

[0080] In the embodiments of the present disclosure, the audio is input into an automatic speech recognition model to be converted into the corresponding text, and then the text is semantically compared with the preset command words to obtain a matching result. The words in the matching result are determined as command words, and semantic comparison is used to improve the accuracy of command word recognition.

[0081] On the basis of the above implementation, each word in the text is marked with the pronunciation time in the audio; determining whether the first audio segment before the target audio segment and / or the second audio segment after the target audio segment in the audio is issued by the user who issues the target audio segment includes: according to the pronunciation time corresponding to each word of the command word in the text, locating the start time point and end time point of the target audio segment in the audio; taking the audio with a preset duration before the start time point as the first audio segment; taking the audio with a preset duration after the end time point as the second audio segment; respectively performing voiceprint feature comparison between the first audio segment and the second audio segment and the target audio segment to obtain the corresponding comparison results; and determining whether the first audio segment and / or the second audio segment is issued by the user who issues the target audio segment according to the corresponding comparison results.

[0082] When the central control computer 205 uses the automatic speech recognition model to convert the collected audio into text, the automatic speech recognition model will calculate the start and end times of each word recognized in the audio through the position of the frames in the audio, so as to obtain the pronunciation time of each word, and mark the pronunciation time corresponding to each word in the text. The central control computer 205 determines the start time point and end time point of the target audio segment in the audio by using the pronunciation time of each word marked in the text. For example, if the pronunciation start time of the word "go" in the command word "<go to the sixth floor>" is the 7th second of the audio, and the pronunciation end time of the word "floor" is the 8th second of the audio, it can be determined that the start time of the target audio segment is the 7th second of the audio, and the end time is the 8th second of the audio.

[0083] The central control computer 205 uses the start time point and the end time point to intercept the audio to obtain the first audio segment and the second audio segment. Continuing with the example above, if the preset duration is set to 3 seconds, the central control computer 205 intercepts the 3-second audio segment before the 7th second of the audio as the first audio segment, that is, the audio segment from the 4th to the 6th second in the audio. Similarly, the second audio segment is the audio segment from the 9th to the 11th second in the audio.

[0084] The central control computer 205 then compares the intercepted first and second audio segments with the target audio segment for voiceprint features, obtains corresponding comparison results, and determines based on the comparison results whether the first and / or second audio segments are uttered by the user who uttered the target audio segment. A voiceprint is a sound wave spectrum that carries speech information, as displayed by electroacoustic instruments. Modern scientific research shows that voiceprints are not only specific but also relatively stable. After adulthood, a person's voice remains relatively stable for a long time. Therefore, by comparing the voiceprint features of audio segments, it is possible to effectively determine whether these audio segments are uttered by the same person.

[0085] In this case, the first audio segment and the second audio segment are respectively compared with the target audio segment for voiceprint features to obtain corresponding comparison results, including: inputting the first audio segment, the second audio segment and the target audio into the voiceprint feature extraction model respectively to obtain the first voiceprint feature, the second voiceprint feature and the third voiceprint feature respectively; determining the first similarity between the first voiceprint feature and the third voiceprint feature, and the second similarity between the second voiceprint feature and the third voiceprint feature; and using the first similarity and the second similarity as the comparison results.

[0086] A voiceprint feature extraction model is a model that converts voiceprint features contained in audio into feature vectors. Most of them are deep learning models. Mainstream voiceprint models include Emphasized Channel Attention, Propagation and Aggregation in time delay neural network Based Speaker Verification (ECAPA-TDNN).

[0087] The central control computer 205 inputs the first audio segment, the second audio segment and the target audio into the voiceprint feature extraction model for feature extraction, obtains the corresponding output first voiceprint feature, second voiceprint feature and third voiceprint feature, and then uses vector similarity calculation to obtain a first similarity between the first voiceprint feature and the third voiceprint feature, and a second similarity between the second voiceprint feature and the third voiceprint feature. The first similarity is the voiceprint feature comparison result between the first audio segment and the target audio segment, and the second similarity is the voiceprint feature comparison result between the second audio segment and the target audio segment.

[0088] Correspondingly, whether the first audio segment and / or the second audio segment is sent by the user who sent the target audio segment is determined based on the corresponding comparison results, including: if the first similarity is greater than the preset threshold, it is determined that the first audio segment is sent by the user who sent the target audio segment; if the second similarity is greater than the preset threshold, it is determined that the second audio segment is sent by the user who sent the target audio segment.

[0089] The central control computer 205 determines whether the first audio segment and / or the second audio segment were uttered by the user who uttered the target audio segment by determining whether the first similarity and the second similarity are greater than a preset threshold. If the first similarity is greater than the preset threshold, the first audio segment is uttered by the user who uttered the target audio segment. If the second similarity is greater than the preset threshold, the second audio segment is uttered by the user who uttered the target audio segment. In other words, when both the first similarity and the second similarity are less than the preset threshold, it can be determined that neither the first audio segment nor the second audio segment was uttered by the user who uttered the target audio segment, and the recognized command word is therefore valid.

[0090] In the disclosed embodiment, the start and end time points of a target audio segment are determined based on the pronunciation time of each character of a command word annotated in the text. A first audio segment and a second audio segment of preset duration are extracted from the audio based on the start and end time points. The first audio segment, the second audio segment, and the target audio segment are then output to a voiceprint feature extraction model to obtain corresponding first, second, and third voiceprint features. A first similarity between the first and third voiceprint features, and a second similarity between the second and third voiceprint features, are then determined. If the first similarity is greater than a preset threshold, the first audio segment and the target audio segment are determined to be from the same user. If the second similarity is greater than a preset threshold, the second audio segment and the target audio segment are determined to be from the same user. This allows determining whether the first audio segment preceding the target audio segment and / or the second audio segment following the target audio segment in the collected audio are from the user who issued the target audio segment, thereby determining whether the user's intention to board the elevator when issuing the command word is credible, and further determining whether the command word is valid.

[0091] Figure 3 This is a schematic diagram of the structure of the voice command recognition device of the intelligent elevator provided by the embodiment of the present disclosure. The voice command recognition device of the intelligent elevator can be a component or assembly in the terminal of the above embodiment. The voice command recognition device of the intelligent elevator provided by the embodiment of the present disclosure can execute the processing flow provided by the embodiment of the voice command recognition method of the intelligent elevator, such as Figure 3 As shown, the voice command recognition device 300 for the intelligent elevator includes: a collection module 301, which is used to collect audio emitted by users in the elevator car; an identification module 302, which is used to identify command words in the audio; a determination module 303, which is used to determine whether the first audio segment before the target audio segment and / or the second audio segment after the target audio segment in the audio are emitted by the user who emitted the target audio segment, where the target audio segment is an audio segment containing a command word in the audio; and an execution module 304, which is used to determine whether the command word is valid and execute the corresponding elevator command according to the command word.

[0092] In some embodiments, the voice command recognition device 300 of the intelligent elevator further includes a noise reduction module 305 for performing noise reduction processing on the audio before recognizing the command words in the audio.

[0093] In some embodiments, when recognizing command words in audio, the recognition module 302 is specifically used to convert the audio into corresponding text; perform semantic comparison between the text and preset command words, and determine that the words in the text that match the preset command words are command words.

[0094] In some embodiments, when converting audio into corresponding text, the recognition module 302 is specifically configured to input the audio into an automatic speech recognition model to obtain text output by the automatic speech recognition model.

[0095] In some embodiments, the text is marked with the pronunciation time of each word in the audio; when determining whether the first audio segment before the target audio segment and / or the second audio segment after the target audio segment in the audio is issued by the user who issued the target audio segment, the determination module 303 is specifically used to locate the starting time point and the ending time point of the target audio segment in the audio according to the pronunciation time corresponding to each word of the command word in the text; take the audio of a preset length before the starting time point as the first audio segment; take the audio of a preset length after the end time point as the second audio segment; perform voiceprint feature comparison on the first audio segment and the second audio segment with the target audio segment respectively to obtain corresponding comparison results; and determine whether the first audio segment and / or the second audio segment is issued by the user who issued the target audio segment based on the corresponding comparison results.

[0096] In some embodiments, when the determination module 303 compares the voiceprint features of the first audio segment and the second audio segment with the target audio segment respectively to obtain corresponding comparison results, it is specifically used to input the first audio segment, the second audio segment and the target audio into the voiceprint feature extraction model respectively, and correspondingly obtain the first voiceprint feature, the second voiceprint feature and the third voiceprint feature; determine the first similarity between the first voiceprint feature and the third voiceprint feature, and the second similarity between the second voiceprint feature and the third voiceprint feature; and use the first similarity and the second similarity as the comparison results.

[0097] In some embodiments, when determining whether the first audio segment and / or the second audio segment is emitted by the user who emitted the target audio segment based on the corresponding comparison results, the determination module 303 is specifically used to determine that the first audio segment is emitted by the user who emitted the target audio segment if the first similarity is greater than a preset threshold; and to determine that the second audio segment is emitted by the user who emitted the target audio segment if the second similarity is greater than the preset threshold.

[0098] Figure 3 The voice command recognition device of the intelligent elevator in the embodiment shown can be used to implement the technical solution of the above-mentioned method embodiment. Its implementation principle and technical effects are similar and will not be repeated here.

[0099] Figure 4 This is a schematic diagram of the structure of the electronic device provided in the embodiment of the present disclosure. The electronic device may be the server as described in the above embodiment. The electronic device provided in the embodiment of the present disclosure may execute the processing flow provided in the embodiment of the voice command recognition method for an intelligent elevator, such as Figure 4 As shown, the electronic device 400 includes: a memory 401, a processor 402, a computer program and a communication interface 403; wherein the computer program is stored in the memory 401 and is configured so that the processor 402 executes the voice command recognition method for the intelligent elevator as described above.

[0100] In addition, an embodiment of the present disclosure further provides a computer-readable storage medium on which a computer program is stored. The computer program is executed by a processor to implement the voice command recognition method for the intelligent elevator described in the above embodiment.

[0101] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0102] The foregoing description is intended only to provide specific embodiments of the present disclosure, intended to enable those skilled in the art to understand and implement the present disclosure. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present disclosure. Therefore, the present disclosure is not intended to be limited to the embodiments described herein, but rather to be construed in the broadest manner consistent with the principles and novel features disclosed herein.

Claims

1. A voice command recognition method for an intelligent elevator, characterized in that: include: Collect audio from users in the elevator car; identifying a command word in the audio; determining whether a first audio segment preceding a target audio segment and / or a second audio segment following the target audio segment in the audio is uttered by a user who uttered the target audio segment, the target audio segment being the audio segment corresponding to the command word in the audio; If not, it is determined that the command word is valid, and the corresponding elevator instruction is executed according to the command word.

2. The method according to claim 1, wherein Before identifying the command word in the audio, the method further includes: Noise reduction processing is performed on the audio.

3. The method according to claim 1, wherein The identifying the command word in the audio includes: Converting the audio into corresponding text; The text is semantically compared with a preset command word, and a word in the text that matches the preset command word is determined to be the command word.

4. The method according to claim 3, wherein The converting the audio into corresponding text includes: The audio is input into an automatic speech recognition model to obtain the text output by the automatic speech recognition model.

5. The method according to claim 4, wherein The text is marked with the pronunciation time of each word in the audio; The determining whether the first audio segment preceding the target audio segment and / or the second audio segment following the target audio segment in the audio is sent by the user who sent the target audio segment includes: Locating the starting time point and the ending time point of the target audio segment in the audio according to the pronunciation time corresponding to each character of the command word in the text; Using the audio of a preset duration before the starting time point as the first audio segment; Using the audio of a preset duration after the end time point as the second audio segment; Comparing the first audio segment and the second audio segment with the target audio segment for voiceprint features, respectively, to obtain corresponding comparison results; Determine whether the first audio segment and / or the second audio segment is uttered by the user who uttered the target audio segment based on the corresponding comparison results.

6. The method according to claim 5, wherein The step of comparing the first audio segment and the second audio segment with the target audio segment respectively to obtain corresponding comparison results includes: Inputting the first audio segment, the second audio segment, and the target audio segment into a voiceprint feature extraction model respectively to obtain a first voiceprint feature, a second voiceprint feature, and a third voiceprint feature respectively; determining a first similarity between the first voiceprint feature and the third voiceprint feature, and a second similarity between the second voiceprint feature and the third voiceprint feature; The first similarity and the second similarity are taken as a comparison result.

7. The method according to claim 6, wherein The determining, based on the corresponding comparison results, whether the first audio segment and / or the second audio segment is sent by the user who sent the target audio segment includes: If the first similarity is greater than a preset threshold, determining that the first audio segment is sent by the user who sent the target audio segment; If the second similarity is greater than a preset threshold, it is determined that the second audio segment is emitted by the user who emitted the target audio segment.

8. A voice command recognition device for an intelligent elevator, characterized in that: include: A collection module, used to collect audio emitted by users in the elevator car; A recognition module, configured to recognize command words in the audio; a determination module, configured to determine whether a first audio segment preceding a target audio segment and / or a second audio segment following the target audio segment in the audio is uttered by a user who uttered the target audio segment, wherein the target audio segment is the audio segment corresponding to the command word in the audio; and an execution module, configured to, upon determining that neither a first audio segment preceding a target audio segment nor a second audio segment following the target audio segment in the audio is issued by the user who issued the target audio segment, determine that the command word is valid, and execute a corresponding elevator instruction according to the command word.

9. An electronic device, characterized in that: include: Memory; processor; as well as computer programs; The computer program is stored in the memory and configured to be executed by the processor to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images