Voice keyword recognition method and system

Abstract

The invention discloses a voice keyword recognition method and system, and relates to the technical field of voice keyword recognition, and the method comprises the steps: 1, voice collection and processing, 2, voice matching, and 3, voice recognition. In the first step of voice acquisition and processing, the voice signal of the user of the intelligent voice assistant is acquired through the voice input device, and the voice recognition coincidence index of the intelligent voice assistant is analyzed, so that the problem of incomplete voice signal recognition is avoided, the recognition efficiency is improved, keywords in the voice of the user are quickly and accurately recognized, and the user experience is improved. The user experience is improved, and the requirements of users are met. In the step 2, in voice matching, the voice decoding coincidence index of the intelligent voice assistant is analyzed, and the quality of voice signals is improved, so that voice recognition errors are reduced, the complex efficiency of voice recognition is reduced, the use difficulty is reduced, and the popularization rate of the intelligent voice assistant is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of voice keywords, and in particular to a method and system for recognizing voice keywords. Background Art

[0002] With the continuous development of speech recognition technology, initially, speech keyword recognition mainly relied on simple template matching and statistical methods, and the recognition effect was limited. However, in recent years, with the rise of deep learning technology, the accuracy of speech keyword recognition will be seriously affected. Therefore, a keyword recognition method is necessary.

[0003] Prior art, such as the invention application patent with announcement number: CN110600058A, discloses a method, device, computer equipment and storage medium based on ultrasonic wake-up of a voice assistant, the method comprising: receiving an ultrasonic wake-up instruction from a wake-up device; analyzing the received ultrasonic wake-up instruction to obtain a spectrum analysis result; comparing the spectrum analysis result with the spectrum of a preset ultrasonic wake-up instruction to determine whether they match; if they match, waking up and activating the voice assistant, and returning confirmation activation information. In this solution, the intelligent voice device collects ultrasonic audio in standby mode, which can prevent the intelligent voice device from listening to user conversations in real time while in standby mode, thereby protecting user privacy; in addition, the difficulty of ultrasonic audio recognition technology is controllable, and compared with the existing wake-up keyword recognition through statistical models or deep neural network models, the requirements for CPU performance and memory capacity are greatly reduced, while ensuring recognition accuracy and efficiency, while reducing the setting cost of the recognition device.

[0004] Based on the above solutions, it can be found that the online gate status monitoring and control in the existing technology can meet the basic requirements, but there are also some potential defects and challenges, which are specifically reflected in the following aspects: 1. The research on the speech recognition compliance index of intelligent voice assistants in the prior art is not in-depth enough. Due to the inaccurate speech signal processing after filtering, frame-segmented speech signal processing, pre-emphasized speech signal processing and endpoint detection, the speech signal recognition is imperfect, which reduces the recognition efficiency, resulting in the inability to quickly and accurately recognize keywords in the user's voice, reducing the user experience and failing to meet user needs.

[0005] 2. The existing technology does not pay enough attention to the voice decoding compliance index of intelligent voice assistants. Due to the inaccuracy of the acoustic model matching data and the language model decoding data of the intelligent voice assistant, the quality of the voice signal is affected, resulting in recognition errors, which increases the complexity and efficiency of voice recognition, and further increases the difficulty of use, affecting the popularity of intelligent voice assistants. Summary of the invention

[0006] The purpose of the present invention is to provide a method and system for recognizing voice keywords, which solves the problems existing in the background technology.

[0007] To solve the above technical problems, the present invention adopts the following technical solutions: The present invention provides a method and system for recognizing voice keywords, which includes step one, voice collection and processing, step two, voice matching and step three, voice recognition.

[0008] Step 1: Voice collection and processing: The voice signal of the user of the intelligent voice assistant is collected through the voice input device, and the collected voice signal is subjected to filtered voice signal processing, framed voice signal processing, pre-emphasized voice signal processing and endpoint detection to obtain the voice signal of the intelligent voice assistant, and then the voice recognition compliance index of the intelligent voice assistant is analyzed.

[0009] Step 2, voice matching: Based on the obtained voice recognition matching index of the intelligent voice assistant, extract the sound feature vector and phrase, obtain the acoustic model matching data of the intelligent voice assistant and the language model decoding data of the intelligent voice assistant, and then analyze the voice decoding matching index of the intelligent voice assistant.

[0010] Step 3: Speech recognition: Based on the speech decoding compliance index of the intelligent voice assistant, detect whether specific keywords exist in the decoded text, and perform corresponding operations based on the detection results.

[0011] Furthermore, the speech recognition compliance index of the intelligent voice assistant is specifically analyzed by: based on the obtained speech signal of the intelligent voice assistant, the front-end signal-to-noise ratio, the back-end signal-to-noise ratio, the short-time average zero-crossing rate, the high-frequency resolution, the starting point speech signal duration and the ending point speech signal duration of the speech signal are extracted to analyze the speech signal noise reduction compliance index of the intelligent voice assistant. The voice processing of the smart voice assistant is consistent with the index , and then analyze the voice recognition compliance index of the intelligent voice assistant. The specific calculation formula is: .

[0012] Furthermore, the voice signal noise reduction compliance index of the intelligent voice assistant is specifically analyzed by the following method: based on the obtained front-end signal-to-noise ratio, rear-end signal-to-noise ratio, short-time average zero-crossing rate and high-frequency resolution of the voice signal of the intelligent voice assistant, and extracting the short-time reference average zero-crossing rate and reference high-frequency resolution of the voice signal of the intelligent voice assistant from the database, the voice signal noise reduction compliance index of the intelligent voice assistant is analyzed, and the specific calculation formula is: ,in It is expressed as the front-end signal-to-noise ratio of the voice signal of the intelligent voice assistant. It is expressed as the signal-to-noise ratio of the back-end of the voice signal of the intelligent voice assistant. It is expressed as the short-time average zero-crossing rate of the voice signal of the intelligent voice assistant. It is represented as the short-term reference average zero-crossing rate of the voice signal of the intelligent voice assistant. Represented as the high-frequency resolution of the voice signal of the intelligent voice assistant, Represents the reference high-frequency resolution of the speech signal for the smart voice assistant.

[0013] Furthermore, the speech processing compliance index of the intelligent voice assistant is specifically analyzed as follows: based on the obtained starting point speech signal duration and ending point speech signal duration of the speech signal of the intelligent voice assistant, the ending point speech signal duration of the speech signal of the intelligent voice assistant is subtracted from the starting point speech signal duration to obtain a valid speech segment, the speech energy of the speech signal of the voice assistant is extracted, and the speech energy of the speech signal of the voice assistant is compared with the reference speech energy qualified interval of the speech signal of the voice assistant stored in the database; if the speech energy of the speech signal of the voice assistant is within the reference speech energy qualified interval, the speech processing compliance index of the intelligent voice assistant is recorded as , otherwise, it is recorded as , and then get the voice processing compliance index of the intelligent voice assistant ,in The value of includes or , .

[0014] Furthermore, the speech decoding conformity index of the intelligent voice assistant is specifically analyzed by the following method: based on the obtained acoustic model matching data of the intelligent voice assistant and the language model decoding data of the intelligent voice assistant, wherein the acoustic model matching data of the intelligent voice assistant includes the phoneme value, speech feature value and annotation restoration value of the speech signal of the intelligent voice assistant, and the language model decoding data of the intelligent voice assistant includes the text conversion rate and vocabulary probability of the speech signal of the intelligent voice assistant, and then the acoustic decoding conformity index of the intelligent voice assistant is analyzed. The language decoding of the voice assistant is consistent with the index , the voice decoding compliance index of the intelligent voice assistant is calculated as follows: .

[0015] Furthermore, the acoustic decoding conformity index of the intelligent voice assistant is specifically analyzed by: based on the obtained phoneme value and voice feature value of the voice signal of the intelligent voice assistant, the phoneme reference value and voice feature reference value of the voice signal of the intelligent voice assistant are extracted from the database, and then the acoustic feature conformity index of the intelligent voice assistant is analyzed, and the specific calculation formula is: ,in Represented as the phoneme value of the voice signal of the intelligent voice assistant, It is represented as the phoneme reference value of the voice signal of the intelligent voice assistant. Represented as the voice feature value of the voice signal of the intelligent voice assistant, It represents the speech feature reference value of the speech signal of the intelligent voice assistant.

[0016] Analyze the acoustic decoding compliance index of the intelligent voice assistant, and its specific calculation formula is: ,in, Expressed as the acoustic conversion compliance index of the smart voice assistant.

[0017] Furthermore, the acoustic conversion conformity index of the intelligent voice assistant is specifically analyzed by: based on the obtained annotated restoration value of the voice signal of the intelligent voice assistant, and extracting the annotated restoration reference value of the voice signal of the intelligent voice assistant from the database, analyzing the acoustic conversion conformity index of the intelligent voice assistant, and the specific calculation formula is: ,in It is represented by the annotated restoration value of the voice signal of the intelligent voice assistant. Represents the reference value of the annotation restoration of the voice signal of the intelligent voice assistant.

[0018] Furthermore, the language decoding conformity index of the voice assistant is specifically analyzed by: based on the text conversion rate and vocabulary probability of the voice signal of the intelligent voice assistant, and extracting the text conversion reference rate and vocabulary reference probability of the voice signal of the intelligent voice assistant from the database, analyzing the language decoding conformity index of the voice assistant, and the specific calculation formula is: , in, The text conversion rate of the voice signal of the intelligent voice assistant is expressed as, It is expressed as the text conversion reference rate of the voice signal of the intelligent voice assistant. Represented as the vocabulary probability of the speech signal of the intelligent voice assistant, Lexical reference probabilities for speech signals represented as intelligent voice assistants.

[0019] Furthermore, the corresponding operation is performed according to the detection result, and its specific analysis method is: based on the obtained speech decoding compliance index of the intelligent voice assistant, the speech decoding compliance index of the intelligent voice assistant is compared with the speech decoding compliance index safety interval of the intelligent voice assistant stored in the database. If the speech decoding compliance index of the intelligent voice assistant is within the speech decoding compliance index safety interval, it means that there are specific keywords in the decoded text, and duplicates are removed, similar keywords are merged, and voice answers are given. If the speech decoding compliance index of the intelligent voice assistant is not within the speech decoding compliance index safety interval, it means that there are no specific keywords in the decoded text, and re-identification is performed.

[0020] The second aspect of the present invention provides a system for a method for recognizing voice keywords, characterized in that it includes: a voice collection and processing module: collecting the voice signal of the user of the intelligent voice assistant through a voice input device, performing filtered voice signal processing, framed voice signal processing, pre-emphasized voice signal processing and endpoint detection on the collected voice signal, and then analyzing the voice recognition compliance index of the intelligent voice assistant.

[0021] Voice matching module: Based on the subsequent recognition matching index of the intelligent voice assistant, the sound feature vector and phrase are extracted to obtain the acoustic model matching data of the intelligent voice assistant and the language model decoding data of the intelligent voice assistant, and then the voice decoding matching index of the intelligent voice assistant is analyzed.

[0022] Speech recognition module: Based on the speech decoding compliance index of the intelligent voice assistant, detect whether specific keywords exist in the decoded text, and perform corresponding operations based on the detection results.

[0023] The beneficial effects of the present invention are as follows: in step one, voice collection and processing: the voice signal of the user of the intelligent voice assistant is collected through a voice input device, and the collected voice signal is subjected to filtered voice signal processing, framed voice signal processing, pre-emphasized voice signal processing and endpoint detection, and then the voice recognition compliance index of the intelligent voice assistant is analyzed, thereby avoiding the problem of imperfect voice signal recognition, improving recognition efficiency, and then quickly and accurately identifying keywords in the user's voice, improving user experience, and meeting user needs.

[0024] In step 2, voice matching: based on the subsequent recognition compliance index of the intelligent voice assistant, the sound feature vector and phrase are extracted to obtain the acoustic model matching data of the intelligent voice assistant and the language model decoding data of the intelligent voice assistant, and then the voice decoding compliance index of the intelligent voice assistant is analyzed to improve the quality of the voice signal, thereby reducing voice recognition errors, reducing the complexity of voice recognition, and then reducing the difficulty of use and increasing the popularity of intelligent voice assistants. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0026] Figure 1 The present invention is a schematic flow chart of the steps for implementing the method.

[0027] Figure 2 It is a schematic diagram of the system structure connection of the present invention. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] Reference Figure 1 As shown, the present invention provides a method and system for recognizing voice keywords, including: step one, voice collection and processing, step two, voice matching and step three, voice recognition.

[0030] Step 1: Voice collection and processing: The voice signal of the user of the intelligent voice assistant is collected through the voice input device, and the collected voice signal is subjected to filtered voice signal processing, framed voice signal processing, pre-emphasized voice signal processing and endpoint detection to obtain the voice signal of the intelligent voice assistant, and then the voice recognition compliance index of the intelligent voice assistant is analyzed.

[0031] In a specific embodiment of the present invention, the speech recognition compliance index of the intelligent voice assistant is analyzed by a specific method: based on the obtained speech signal of the intelligent voice assistant, the front-end signal-to-noise ratio, the back-end signal-to-noise ratio, the short-time average zero-crossing rate, the high-frequency resolution, the starting point speech signal duration and the ending point speech signal duration of the speech signal are extracted to analyze the speech signal noise reduction compliance index of the intelligent voice assistant. The voice processing of the smart voice assistant is consistent with the index , and then analyze the voice recognition compliance index of the intelligent voice assistant. The specific calculation formula is: .

[0032] In a specific embodiment of the present invention, the voice signal noise reduction compliance index of the intelligent voice assistant is specifically analyzed by the following method: based on the obtained front-end signal-to-noise ratio, rear-end signal-to-noise ratio, short-time average zero-crossing rate and high-frequency resolution of the voice signal of the intelligent voice assistant, the short-time reference average zero-crossing rate and reference high-frequency resolution of the voice signal of the intelligent voice assistant are extracted from the database, and the voice signal noise reduction compliance index of the intelligent voice assistant is analyzed, and the specific calculation formula is: ,in It is expressed as the front-end signal-to-noise ratio of the voice signal of the intelligent voice assistant. It is expressed as the signal-to-noise ratio of the back-end of the voice signal of the intelligent voice assistant. It is expressed as the short-time average zero-crossing rate of the voice signal of the intelligent voice assistant. It is represented as the short-term reference average zero-crossing rate of the voice signal of the intelligent voice assistant. Represented as the high-frequency resolution of the voice signal of the intelligent voice assistant, Represents the reference high-frequency resolution of the speech signal for the smart voice assistant.

[0033] It should be noted that the short-time reference average zero-crossing rate of the voice signal of the intelligent voice assistant refers to an important parameter used to describe the characteristics of the voice signal, which reflects the average number of times the voice signal passes through the time axis in a historical unit time.

[0034] It should be noted that the reference high-frequency resolution of the voice signal of the intelligent voice assistant refers to the historical minimum frequency difference that the system can distinguish and resolve during the signal processing process of the historical high-frequency resolution.

[0035] In a specific embodiment of the present invention, the speech processing compliance index of the intelligent voice assistant is specifically analyzed by the following method: based on the obtained starting point speech signal duration and ending point speech signal duration of the speech signal of the intelligent voice assistant, the ending point speech signal duration of the speech signal of the intelligent voice assistant is subtracted from the starting point speech signal duration to obtain a valid speech segment, the speech energy of the speech signal of the voice assistant is extracted, and the speech energy of the speech signal of the voice assistant is compared with the reference speech energy qualified interval of the speech signal of the voice assistant stored in the database; if the speech energy of the speech signal of the voice assistant is within the reference speech energy qualified interval, the speech processing compliance index of the intelligent voice assistant is recorded as , otherwise, it is recorded as , and then get the voice processing compliance index of the intelligent voice assistant ,in The value of includes or , .

[0036] It should be noted that the reference speech energy qualified interval refers to a series of standard values ​​or thresholds used to measure whether the speech signal energy is within a reasonable or expected range in speech signal processing.

[0037] It should be noted that in step one, voice collection and processing: the voice signal of the user of the intelligent voice assistant is collected through a voice input device, and the collected voice signal is subjected to filtered voice signal processing, framed voice signal processing, pre-emphasized voice signal processing and endpoint detection, and then the voice recognition compliance index of the intelligent voice assistant is analyzed, thereby avoiding the problem of imperfect voice signal recognition, improving recognition efficiency, and then quickly and accurately identifying keywords in the user's voice, improving user experience, and meeting user needs.

[0038] Step 2, voice matching: Based on the obtained voice recognition matching index of the intelligent voice assistant, extract the sound feature vector and phrase, obtain the acoustic model matching data of the intelligent voice assistant and the language model decoding data of the intelligent voice assistant, and then analyze the voice decoding matching index of the intelligent voice assistant.

[0039] In a specific embodiment of the present invention, the speech decoding conformity index of the intelligent voice assistant is specifically analyzed by the following method: based on the obtained acoustic model matching data of the intelligent voice assistant and the language model decoding data of the intelligent voice assistant, wherein the acoustic model matching data of the intelligent voice assistant includes the phoneme value, speech feature value and annotation restoration value of the speech signal of the intelligent voice assistant, and the language model decoding data of the intelligent voice assistant includes the text conversion rate and vocabulary probability of the speech signal of the intelligent voice assistant, and then the acoustic decoding conformity index of the intelligent voice assistant is analyzed. The language decoding of the voice assistant is consistent with the index , the voice decoding compliance index of the intelligent voice assistant is calculated as follows: .

[0040] In a specific embodiment of the present invention, the acoustic decoding conformity index of the intelligent voice assistant is specifically analyzed by: based on the obtained phoneme value and voice feature value of the voice signal of the intelligent voice assistant, the phoneme reference value and voice feature reference value of the voice signal of the intelligent voice assistant are extracted from the database, and then the acoustic feature conformity index of the intelligent voice assistant is analyzed, and the specific calculation formula is: ,in Represented as the phoneme value of the voice signal of the intelligent voice assistant, It is represented as the phoneme reference value of the voice signal of the intelligent voice assistant. Represented as the voice feature value of the voice signal of the intelligent voice assistant, It represents the speech feature reference value of the speech signal of the intelligent voice assistant.

[0041] It should be noted that the phoneme reference value of the voice signal of the intelligent voice assistant refers to the characteristic value used to describe and identify the basic pronunciation unit in the voice when processing and analyzing the voice signal.

[0042] It should be noted that the speech feature reference value of the speech signal of the intelligent voice assistant refers to the standard parameter used to describe and quantify the characteristics of the speech signal when processing and analyzing the speech signal.

[0043] Analyze the acoustic decoding compliance index of the intelligent voice assistant, and its specific calculation formula is: ,in, Expressed as the acoustic conversion compliance index of the smart voice assistant.

[0044] In a specific embodiment of the present invention, the acoustic conversion conformity index of the intelligent voice assistant is specifically analyzed by: based on the obtained annotated restoration value of the voice signal of the intelligent voice assistant, and extracting the annotated restoration reference value of the voice signal of the intelligent voice assistant from the database, analyzing the acoustic conversion conformity index of the intelligent voice assistant, and the specific calculation formula is: ,in It is represented by the annotated restoration value of the voice signal of the intelligent voice assistant. Represents the reference value of the annotation restoration of the voice signal of the intelligent voice assistant.

[0045] It should be noted that the labeled restoration reference value of the voice signal of the intelligent voice assistant refers to the numerical value used to evaluate and optimize the performance of the speech recognition or synthesis system when processing and analyzing the voice signal.

[0046] In a specific embodiment of the present invention, the language decoding conformity index of the voice assistant is specifically analyzed by: based on the text conversion rate and vocabulary probability of the voice signal of the intelligent voice assistant obtained, and extracting the text conversion reference rate and vocabulary reference probability of the voice signal of the intelligent voice assistant from the database, analyzing the language decoding conformity index of the voice assistant, and the specific calculation formula is: , in, The text conversion rate of the voice signal of the intelligent voice assistant is expressed as, It is expressed as the text conversion reference rate of the voice signal of the intelligent voice assistant. Represented as the vocabulary probability of the speech signal of the intelligent voice assistant, Lexical reference probabilities for speech signals represented as intelligent voice assistants.

[0047] It should be noted that the text-to-text conversion reference rate of the intelligent voice assistant's voice signal refers to the proportion of the voice signal that the system correctly recognizes and converts into text during the voice recognition process. It is calculated by comparing the difference between the text output by the system and the actual recording or annotated text.

[0048] It should be noted that the vocabulary reference probability of the voice signal of the intelligent voice assistant refers to the numerical value used to describe and quantify the possibility or probability of the occurrence of specific words when processing and analyzing voice signals.

[0049] It should be noted that in step 2, voice matching: based on the subsequent recognition compliance index of the intelligent voice assistant obtained, the sound feature vectors and phrases are extracted, the acoustic model matching data of the intelligent voice assistant and the language model decoding data of the intelligent voice assistant are obtained, and then the voice decoding compliance index of the intelligent voice assistant is analyzed, which improves the quality of the voice signal, thereby reducing voice recognition errors, reducing the complexity and efficiency of voice recognition, and then reducing the difficulty of use and increasing the popularity of intelligent voice assistants.

[0050] Step 3: Speech recognition: Based on the speech decoding compliance index of the intelligent voice assistant, detect whether specific keywords exist in the decoded text, and perform corresponding operations based on the detection results.

[0051] In a specific embodiment of the present invention, the corresponding operation is performed according to the detection result, and the specific analysis method is: based on the obtained speech decoding compliance index of the intelligent voice assistant, the speech decoding compliance index of the intelligent voice assistant is compared with the speech decoding compliance index safety interval of the intelligent voice assistant stored in the database. If the speech decoding compliance index of the intelligent voice assistant is within the speech decoding compliance index safety interval, it means that there are specific keywords in the decoded text, and duplicates are removed, similar keywords are merged, and voice answers are given. If the speech decoding compliance index of the intelligent voice assistant is not within the speech decoding compliance index safety interval, it means that there are no specific keywords in the decoded text, and re-identification is performed.

[0052] Reference Figure 2 As shown, the present invention provides a system for a method for recognizing voice keywords, characterized in that it includes: a voice collection and processing module: collecting the voice signal of the user of the intelligent voice assistant through a voice input device, performing filtered voice signal processing, framed voice signal processing, pre-emphasized voice signal processing and endpoint detection on the collected voice signal, and then analyzing the voice recognition compliance index of the intelligent voice assistant.

[0053] Voice matching module: Based on the subsequent recognition matching index of the intelligent voice assistant, the sound feature vector and phrase are extracted to obtain the acoustic model matching data of the intelligent voice assistant and the language model decoding data of the intelligent voice assistant, and then the voice decoding matching index of the intelligent voice assistant is analyzed.

[0054] Speech recognition module: Based on the speech decoding compliance index of the intelligent voice assistant, detect whether specific keywords exist in the decoded text, and perform corresponding operations based on the detection results.

[0055] It should be noted that the database is used to store the short-time reference average zero-crossing rate, reference high-frequency resolution, reference speech energy qualified range, phoneme reference value and speech feature reference value of the intelligent voice assistant's voice signal, annotation and restoration reference value of the voice signal, text conversion reference rate, vocabulary reference probability and speech decoding compliance index safety range.

[0056] The above contents are merely examples and explanations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, they shall all fall within the protection scope of the present invention.

Claims

1. A method for recognizing speech keywords, characterized in that: include: The voice signal of the user of the intelligent voice assistant is collected through the voice input device, and the collected voice signal is subjected to voice signal processing after filtering, voice signal processing after framing, voice signal processing after pre-emphasis, and endpoint detection to obtain the voice signal of the intelligent voice assistant, and then the voice recognition compliance index of the intelligent voice assistant is analyzed; Based on the obtained speech recognition compliance index of the intelligent voice assistant, the sound feature vector and phrase are extracted to obtain the acoustic model matching data of the intelligent voice assistant and the language model decoding data of the intelligent voice assistant, and then the speech decoding compliance index of the intelligent voice assistant is analyzed; Based on the obtained voice decoding compliance index of the intelligent voice assistant, whether specific keywords exist in the decoded text is detected, and corresponding operations are performed according to the detection results.

2. The method for recognizing a voice keyword according to claim 1, characterized in that: The speech recognition compliance index of the intelligent voice assistant is analyzed in the following specific method: Based on the obtained voice signal of the intelligent voice assistant, the front-end signal-to-noise ratio, the back-end signal-to-noise ratio, the short-time average zero-crossing rate, the high-frequency resolution, the duration of the voice signal at the starting point and the duration of the voice signal at the ending point are extracted to analyze the voice signal noise reduction compliance index of the intelligent voice assistant. The voice processing of the smart voice assistant is consistent with the index , and then analyze the voice recognition compliance index of the intelligent voice assistant. The specific calculation formula is: .

3. A method for recognizing speech keywords according to claim 2, characterized in that: The voice signal noise reduction compliance index of the intelligent voice assistant is analyzed in the following specific method: Based on the obtained front-end signal-to-noise ratio, back-end signal-to-noise ratio, short-time average zero-crossing rate and high-frequency resolution of the voice signal of the intelligent voice assistant, and extracting the short-time reference average zero-crossing rate and reference high-frequency resolution of the voice signal of the intelligent voice assistant from the database, the voice signal noise reduction compliance index of the intelligent voice assistant is analyzed, and its specific calculation formula is: ,in It is expressed as the front-end signal-to-noise ratio of the voice signal of the intelligent voice assistant. It is expressed as the signal-to-noise ratio of the back-end of the voice signal of the intelligent voice assistant. It is expressed as the short-time average zero-crossing rate of the voice signal of the intelligent voice assistant. It is represented as the short-term reference average zero-crossing rate of the voice signal of the intelligent voice assistant. Represented as the high-frequency resolution of the voice signal of the intelligent voice assistant, Represents the reference high-frequency resolution of the speech signal for the smart voice assistant.

4. The method for recognizing a voice keyword according to claim 2, characterized in that: The voice processing compliance index of the intelligent voice assistant is analyzed in the following specific way: Based on the obtained starting point voice signal duration and ending point voice signal duration of the voice signal of the intelligent voice assistant, the ending point voice signal duration of the voice signal of the intelligent voice assistant is subtracted from the starting point voice signal duration to obtain a valid voice segment, and the voice energy of the voice signal of the voice assistant is extracted. The voice energy of the voice signal of the voice assistant is compared with the reference voice energy qualified interval of the voice signal of the voice assistant stored in the database. If the voice energy of the voice signal of the voice assistant is within the reference voice energy qualified interval, the voice processing compliance index of the intelligent voice assistant is recorded as , otherwise, it is recorded as , and then get the voice processing compliance index of the intelligent voice assistant ,in The value of includes or , .

5. The method for recognizing a voice keyword according to claim 1, characterized in that: The voice decoding compliance index of the intelligent voice assistant is analyzed in the following specific method: Based on the obtained acoustic model matching data of the intelligent voice assistant and the language model decoding data of the intelligent voice assistant, wherein the acoustic model matching data of the intelligent voice assistant includes the phoneme value, voice feature value and annotation restoration value of the voice signal of the intelligent voice assistant, and the language model decoding data of the intelligent voice assistant includes the text conversion rate and vocabulary probability of the voice signal of the intelligent voice assistant, the acoustic decoding compliance index of the intelligent voice assistant is analyzed. The language decoding of the voice assistant is consistent with the index , the voice decoding compliance index of the intelligent voice assistant is calculated as follows: .

6. A method for recognizing speech keywords according to claim 5, characterized in that: The acoustic decoding compliance index of the intelligent voice assistant is analyzed in the following specific method: Based on the obtained phoneme value and voice feature value of the voice signal of the intelligent voice assistant, the phoneme reference value and voice feature reference value of the voice signal of the intelligent voice assistant are extracted from the database, and then the acoustic feature conformity index of the intelligent voice assistant is analyzed. The specific calculation formula is: ,in Represented as the phoneme value of the voice signal of the intelligent voice assistant, It is represented as the phoneme reference value of the voice signal of the intelligent voice assistant. Represented as the voice feature value of the voice signal of the intelligent voice assistant, The voice feature reference value of the voice signal represented as an intelligent voice assistant; Analyze the acoustic decoding compliance index of the intelligent voice assistant, and its specific calculation formula is: ,in, Expressed as the acoustic conversion compliance index of the smart voice assistant.

7. A method for recognizing speech keywords according to claim 6, characterized in that: The acoustic conversion compliance index of the intelligent voice assistant is analyzed in the following specific method: Based on the obtained annotated restoration value of the voice signal of the intelligent voice assistant, the annotated restoration reference value of the voice signal of the intelligent voice assistant is extracted from the database, and the acoustic conversion compliance index of the intelligent voice assistant is analyzed. The specific calculation formula is: ,in It is represented by the annotated restoration value of the voice signal of the intelligent voice assistant. Represents the reference value of the annotation restoration of the voice signal of the intelligent voice assistant.

8. The method for recognizing a voice keyword according to claim 5, characterized in that: The language decoding compliance index of the voice assistant is analyzed in the following specific method: Based on the obtained text conversion rate and vocabulary probability of the voice signal of the intelligent voice assistant, the text conversion reference rate and vocabulary reference probability of the voice signal of the intelligent voice assistant are extracted from the database, and the language decoding compliance index of the voice assistant is analyzed. The specific calculation formula is: , in, The text conversion rate of the voice signal of the intelligent voice assistant is expressed as, It is expressed as the text conversion reference rate of the voice signal of the intelligent voice assistant. Represented as the vocabulary probability of the speech signal of the intelligent voice assistant, Lexical reference probabilities for speech signals represented as intelligent voice assistants.

9. The method for recognizing a voice keyword according to claim 1, characterized in that: The corresponding operation is performed according to the detection result, and the specific analysis method is as follows: Based on the obtained speech decoding compliance index of the intelligent voice assistant, the speech decoding compliance index of the intelligent voice assistant is compared with the speech decoding compliance index safety interval of the intelligent voice assistant stored in the database. If the speech decoding compliance index of the intelligent voice assistant is within the speech decoding compliance index safety interval, it means that there are specific keywords in the decoded text, and duplicates are removed, similar keywords are merged, and voice answers are given. If the speech decoding compliance index of the intelligent voice assistant is not within the speech decoding compliance index safety interval, it means that there are no specific keywords in the decoded text, and re-recognition is performed.

10. A system for executing the method for recognizing speech keywords according to any one of claims 1 to 9, characterized in that: include: Voice collection and processing module: collects the voice signals of the users of the intelligent voice assistant through the voice input device, performs voice signal processing after filtering, voice signal processing of framing, voice signal processing of pre-emphasis and endpoint detection on the collected voice signals, and then analyzes the voice recognition compliance index of the intelligent voice assistant; Voice matching module: based on the subsequent recognition matching index of the intelligent voice assistant, extract the sound feature vector and phrase, obtain the acoustic model matching data of the intelligent voice assistant and the language model decoding data of the intelligent voice assistant, and then analyze the voice decoding matching index of the intelligent voice assistant; Speech recognition module: Based on the speech decoding compliance index of the intelligent voice assistant, detect whether specific keywords exist in the decoded text, and perform corresponding operations based on the detection results.

Images