A human-machine interaction method and system for speech recognition

Abstract

The present application relates to the technical field of speech recognition, and particularly relates to a human-computer interaction method and system for speech recognition, which comprises the following steps: obtaining the spectral information of a user speech segment to generate a spectrogram, constructing a user voice library, and performing word segmentation to obtain a sequence of word units; using a clustering algorithm to divide word groups, calculating the segmentation coefficient and the stop coefficient of each word group according to the mapping function of the upper and lower envelope lines of the spectrogram, so as to determine the optimal clustering result; comparing the real-time input speech segment with the reference word group, calculating the user confirmation coefficient between the spectrogram of the real-time word group and the spectrogram of the reference word group; and judging whether the real-time input speech segment and the reference speech segment are the same user according to the user confirmation coefficient, thereby completing the human-computer interaction process of speech recognition. The present application constructs a user voice library, uses spectrogram analysis and a clustering algorithm, realizes efficient speech recognition and user confirmation, and improves the user interaction experience.

Description

Technical Field

[0001] The present invention relates to the field of speech recognition technology, and in particular to a human-computer interaction method and system for speech recognition. Background Art

[0002] Speech recognition is a technology that converts human speech signals into text or commands, enabling computers or other devices to understand and process human voice input. By analyzing the acoustic characteristics of speech signals, speech recognition systems can identify words or commands in speech and convert them into actionable information. The core of this technology lies in adapting to the language habits and speech characteristics of different users, thereby achieving efficient human-computer interaction.

[0003] Speech recognition technology plays a crucial role in human-computer interaction. It not only improves the efficiency and convenience of interaction, but also enhances the usability and intelligence of devices. In areas such as smart homes, smart assistants, and in-vehicle systems, speech recognition enables users to communicate with devices through natural language, eliminating manual operation and achieving a seamless interactive experience. Furthermore, speech recognition provides a more user-friendly interaction method for people with disabilities, further promoting the widespread adoption and application of this technology.

[0004] Despite significant progress in speech recognition technology, practical applications still face several challenges. First, speech recognition systems struggle to fully account for each user's unique language habits and speech characteristics when processing speech from different users. For example, changes in the emotional state of a sentence can lead to significant differences in the pronunciation of the same word, affecting the system's accuracy. Second, identity recognition methods based on fixed speech templates struggle to adapt to these variations, reducing the security and reliability of human-computer interaction. Furthermore, existing technologies remain deficient in extracting and utilizing key information from speech, further limiting the accuracy and practicality of speech recognition. Summary of the Invention

[0005] To address the problem that existing speech recognition technology is difficult to adapt to individual language habits and voice feature differences, resulting in low identity recognition accuracy and insufficient extraction of key information, which affects the security and reliability of human-computer interaction, the present invention provides solutions in the following aspects.

[0006] In a first aspect, a human-computer interaction method for speech recognition includes: obtaining spectral information of several historical speech segments of a user, generating a spectrogram using Fourier transform, building a speech library of the user, performing word segmentation processing on each speech segment in the speech library to obtain a sequence of word units; dividing the sequence of word units using a clustering algorithm to obtain word groups consisting of multiple continuous word units, using each word group in the speech library as a reference word group, and using the spectrogram corresponding to each word group as a reference speech segment; the clustering process includes: calculating the segmentation coefficient of each word group according to the mapping function of the upper and lower envelopes of the spectrogram, setting the initial number of clusters, According to the segmentation coefficient, the word units are allocated to the clusters in order. According to the number of word units and the segmentation coefficient, the stopping coefficient of the cluster division is constructed, and clustering iteration is performed, and the best clustering result is output based on the stopping coefficient; the spectrum information of the real-time input voice segment and the real-time word group after division are obtained, and in response to the real-time word group being the same as the reference word group, the spectrogram of the real-time word group is retained, and the similarity between the spectrogram of the real-time word group and the spectrogram of the reference word group is used as the user confirmation coefficient; according to the user confirmation coefficient, it is judged whether the real-time input voice segment and the reference voice segment are from the same user, and the human-computer interaction process of voice recognition is completed.

[0007] By building a user's voice database and reference word groups, using Fourier transforms to generate spectrograms and clustering algorithms to segment word groups, the system can more accurately recognize user voice input. Adapting to different users' language habits and voice characteristics, the system dynamically adjusts clustering results by analyzing the upper and lower envelopes of the spectrogram and the segmentation coefficient to ensure reliable recognition. Users can complete identity verification with just voice input, eliminating the need to enter a password or use other biometric identification methods. This simplifies the verification process, improves system usability, and enhances user satisfaction. Furthermore, by calculating the user confirmation coefficient, the system can determine whether the real-time input voice clip and the reference voice clip are from the same user, enhancing system security and preventing unauthorized access. Furthermore, by optimizing the clustering process to avoid over-segmentation, system performance is improved. The system can process user voice input in real time and quickly output recognition results, enhancing system responsiveness and user interaction.

[0008] Preferably, the word segmentation processing of each speech segment in the speech database to obtain a sequence of word units includes:

[0009] Convert each speech sentence in the speech library to text to obtain the corresponding text sentence. Use the jieba library to divide the text sentence into word units to obtain a sequence of word units, where a text sentence contains several consecutive word units.

[0010] By converting spoken sentences from a speech library into text and using the Jieba library for word segmentation, we can break down continuous text sentences into multiple word units, forming sequences of these units. This not only improves speech recognition accuracy but also enhances text processing efficiency and system adaptability, enabling better understanding and processing of various language structures and semantic information.

[0011] Preferably, the word unit is the basic component of speech or text, and is used to represent the smallest semantic unit of speech signal or text content; the word group is a set composed of several consecutive word units, and is used to represent a specific part of speech or text; the text sentence is the complete text content after speech transcription, and is a continuous text.

[0012] Preferably, the division coefficient includes:

[0013] Integrate the area between adjacent minimum points of the upper envelope, calculate the signal energy in the area between adjacent minimum points, divide the integration result by the product of the interval width and the maximum value between adjacent minimum points, and use it to normalize the integration result. Sum the normalized results of all adjacent minimum values ​​and divide it by the number of minimum points to obtain the mean value of the upper envelope;

[0014] Integrate the area between adjacent maximum points of the lower envelope, calculate the signal energy in the area between adjacent maximum points, divide the integration result by the product of the interval width and the minimum value between adjacent maximum points, and use it to normalize the integration result. Sum the normalized results of all adjacent maxima and divide by the number of maximum points to obtain the mean of the lower envelope;

[0015] The average value of the sum of the mean of the upper envelope and the mean of the lower envelope is subtracted from 1 to obtain the segmentation coefficient of each word group.

[0016] By comprehensively analyzing the upper and lower envelope features of the speech signal, it is possible to capture the high-frequency and low-frequency changes in the speech signal, helping the system to more accurately identify subtle differences in speech, reduce misrecognition, adapt to the speech habits and emotional changes of different users, and ensure that the system can still provide reliable recognition results under variable speech input.

[0017] Preferably, the stop coefficient includes:

[0018] Calculate the ratio of the number of word units in each word group to the total number of word units in the entire text sentence to obtain the contribution degree of each word group;

[0019] The average value of the product of the contribution degree of all word groups and the segmentation coefficient is used as the stopping coefficient of the corresponding cluster division.

[0020] By comprehensively considering the contribution and feature significance of each word group, the quality of the clustering results can be evaluated more accurately, thereby determining the optimal number of clusters, improving the accuracy and efficiency of clustering, and providing a more reliable foundation for speech recognition and human-computer interaction.

[0021] Preferably, outputting the best clustering result based on the stopping coefficient includes:

[0022] In response to the stop coefficient being greater than the preset stop threshold, or when the stop coefficient reaches the maximum value, the output clustering result is the optimal clustering result, wherein in the optimal clustering result, each cluster corresponds to a word group, and the number of the clusters does not exceed the number of word groups.

[0023] By dynamically determining the optimal number of clusters, over-segmentation is avoided, ensuring the accuracy and efficiency of the clustering results, while improving the adaptability and robustness of the system, providing reliable support for speech recognition and human-computer interaction.

[0024] Preferably, the user confirmation coefficient includes:

[0025] Calculate the absolute difference between the time length of the spectrogram of the reference word group and the time length of the spectrogram of the real-time word group, and normalize the difference based on the ratio of the absolute difference to the reference time length. The difference between 1 and the normalized ratio is used to obtain the degree of match in time length, which is used to measure the similarity in time length between the reference word group and the real-time word group.

[0026] Using the DTW algorithm, the similarity between the frequency sequence of the spectrogram of the reference word group and the frequency sequence of the spectrogram of the real-time word group is calculated, and the ratio between the matching degree and the similarity between the frequency sequences and the sum of the hyperparameters is used as the user confirmation coefficient.

[0027] Preferably, the user confirmation coefficient also includes:

[0028] Calculate the similarity between the frequency sequence of the reference word group and the frequency sequence of the real-time word group using the cosine similarity algorithm;

[0029] Calculate the absolute difference between the time length of the spectrogram of the reference word group and the time length of the spectrogram of the real-time word group, and normalize the difference based on the ratio of the absolute difference to the reference time length. The difference between 1 and the normalized ratio is used to obtain the degree of match in time length, which is used to measure the similarity in time length between the reference word group and the real-time word group.

[0030] The product of the similarity and matching degree between frequency sequences is used as the user confirmation coefficient.

[0031] Preferably, the step of determining whether the voice segment input in real time and the reference voice segment are from the same user based on the user confirmation coefficient includes:

[0032] In response to the user confirmation coefficient being less than or equal to the preset similarity threshold, it is considered that the input voice segment and the reference voice segment are not from the same user; otherwise, it is considered that the input voice segment and the reference voice segment are from the same user.

[0033] In a second aspect, a human-computer interaction system for speech recognition includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the human-computer interaction method for speech recognition is implemented.

[0034] The present invention has the following effects:

[0035] 1. This invention builds a user voice library and reference word groups, uses Fourier transform to generate spectrograms, and uses clustering algorithms to divide word groups. This allows it to capture key features of speech signals and improve speech recognition accuracy. Furthermore, by calculating the user confirmation coefficient, the system can more accurately determine whether the real-time input voice segment and the reference voice segment originate from the same user, further improving recognition accuracy.

[0036] 2. The present invention can adapt to the unique language habits and speech characteristics of different users. By analyzing the upper and lower envelopes and segmentation coefficients of the spectrogram, the system can dynamically adjust the clustering results to ensure the reliability of the recognition results. The use of the jieba library for word segmentation can better understand and process various language structures and semantic information, thereby enhancing the adaptability of the system.

[0037] 3. The present invention reduces the reliance on additional biometric technology or complex password verification and completes identity confirmation through voice input, which not only improves the usability of the system but also enhances user satisfaction with the system. At the same time, by calculating the user confirmation coefficient, the system can determine whether the real-time input voice segment and the reference voice segment are from the same user, thereby improving the security of the system and preventing unauthorized access. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a method flow chart of steps S1 to S4 in a human-computer interaction method for speech recognition according to an embodiment of the present invention.

[0039] Figure 2 This is a structural block diagram of a human-computer interaction system for speech recognition according to an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0041] Reference Figure 1 A human-computer interaction method for speech recognition includes steps S1 to S4, which are specifically as follows:

[0042] S1: Obtain the spectrum information of several historical speech segments of the user, use Fourier transform to generate a spectrogram, build the user's speech library, perform word segmentation on each speech segment in the speech library, and obtain a sequence of word units.

[0043] Convert each speech sentence in the speech library to text to obtain the corresponding text sentence. Use the jieba library to divide the text sentence into word units to obtain a sequence of word units, where a text sentence contains several consecutive word units.

[0044] A word unit is the basic component of speech or text, and is used to represent the smallest semantic unit of speech signals or text content; a word group is a collection of several consecutive word units, used to represent a specific part of speech or text; a text sentence is the complete text content after speech transcription, and is a continuous text.

[0045] For example, take the following sentence: "Speech recognition technology is widely used in smart homes":

[0046] The text sentence is: Voice recognition technology is widely used in smart homes.

[0047] The word groups are: voice recognition technology, in smart home, widely used.

[0048] The word units are: voice, recognition, technology, in, intelligence, home, in, application, and widely.

[0049] S2: Use a clustering algorithm to divide the sequence of word units to obtain word groups composed of multiple consecutive word units, use each word group in the speech library as a reference word group, and use the spectrogram corresponding to each word group as a reference speech segment.

[0050] Among them, the clustering process includes: calculating the segmentation coefficient of each word group according to the mapping function of the upper and lower envelopes of the spectrogram, setting the initial number of clusters, and allocating the word units to the clusters in order according to the segmentation coefficient. According to the number of word units and the segmentation coefficient, the stopping coefficient of the cluster division is constructed, and clustering iteration is performed to output the best clustering result based on the stopping coefficient.

[0051] Split factors, including:

[0052] Integrate the area between adjacent minimum points of the upper envelope, calculate the signal energy in the area between adjacent minimum points, divide the integration result by the product of the interval width and the maximum value between adjacent minimum points, and use it to normalize the integration result. Sum the normalized results of all adjacent minimum values ​​and divide it by the number of minimum points to obtain the mean value of the upper envelope;

[0053] Integrate the area between adjacent maximum points of the lower envelope, calculate the signal energy in the area between adjacent maximum points, divide the integration result by the product of the interval width and the minimum value between adjacent maximum points, and use it to normalize the integration result. Sum the normalized results of all adjacent maxima and divide by the number of maximum points to obtain the mean of the lower envelope;

[0054] The average value of the sum of the mean of the upper envelope and the mean of the lower envelope is subtracted from 1 to obtain the segmentation coefficient of each word group.

[0055] Specifically, the division coefficient satisfies the following relationship:

[0056] ;

[0057] Where, Indicates the The division coefficient of word groups, represents the number of minima of the upper envelope, Indicates the upper envelope A minimum value, Indicates the upper envelope A minimum value, represents the mapping function of the upper envelope, represents the mapping function of the lower envelope, Indicates the upper envelope The horizontal coordinate corresponding to the minimum value is Indicates the upper envelope The minimum and The maximum value between the minimum values, represents the number of maxima of the lower envelope, The lower envelope A maximum value, The lower envelope A maximum value, Indicates the upper envelope The maximum and The minimum value between the maximum values.

[0058] That is to say, Reflects the high amplitude area of ​​the speech signal, Reflects the low-amplitude region of the speech signal;

[0059] The continuous variation of the maximum value of the speech spectrogram can be expressed by taking the average of the ratio of the integral of the two minimum intervals in the upper envelope to the integral of the whole. When this value is small, it means that the continuous variation of the high-amplitude area is large, and the speech segment is more distinctive in the high-amplitude area, and is more likely to have distinctive features.

[0060] The continuous variation of the minimum value of the speech spectrogram can be expressed by taking the average of the ratio of the integral of the two maximum intervals in the lower envelope to the integral of the whole. When this value is small, it means that the continuous variation of the low-amplitude region is large, and the speech segment is more distinctive in the low-amplitude region, which is more likely to have distinctive features.

[0061] The stopping factors include:

[0062] Calculate the ratio of the number of word units in each word group to the total number of word units in the entire text sentence to obtain the contribution degree of each word group;

[0063] The average value of the product of the contribution degree of all word groups and the segmentation coefficient is used as the stopping coefficient of the corresponding cluster division.

[0064] Specifically, the stop coefficient satisfies the following relationship:

[0065] ;

[0066] Where, represents the stopping coefficient, Indicates the number of clusters, Indicates the The ratio of the number of word units contained in the word phrase to the total number of word units in the entire text sentence, Indicates the The division coefficient of each word group.

[0067] In other words, the number of clusters is limited by the proportion of the number of elements in the cluster (the number of word units in the word group) to the entire text sentence. The value growth prevents the stop condition from reaching the maximum value. The value is directly equal to the number of word units in the text sentence. This kind of clustering is meaningless, and the larger the proportion, the longer the word group is, which means that the effect of other text sentences will be more different in subsequent comparisons.

[0068] Output the best clustering results based on the stopping coefficient, including:

[0069] In response to the stop coefficient being greater than the preset stop threshold, or when the stop coefficient reaches the maximum value, the output clustering result is the optimal clustering result, wherein in the optimal clustering result, each cluster corresponds to a word group, and the number of clusters does not exceed the number of word groups.

[0070] It should be noted that the default stopping threshold is 0.6, which can be adjusted based on specific implementation circumstances. When a speech segment contains a large number of word groups, the clustering process may generate a large number of clusters. This not only increases computational complexity but can also lead to over-segmentation, making the features of each cluster less distinct. In this case, using a stopping threshold can effectively control the clustering process and ensure the quality and efficiency of the clustering results.

[0071] S3: Obtain the spectrum information of the real-time input voice segment and the divided real-time word groups. In response to the real-time word group being the same as the reference word group, retain the spectrogram of the real-time word group, and use the similarity between the spectrogram of the real-time word group and the spectrogram of the reference word group as the user confirmation coefficient.

[0072] User confirmation factors include:

[0073] Calculate the absolute difference between the time length of the spectrogram of the reference word group and the time length of the spectrogram of the real-time word group, and normalize the difference based on the ratio of the absolute difference to the reference time length. The difference between 1 and the normalized ratio is used to obtain the degree of match in time length, which is used to measure the similarity in time length between the reference word group and the real-time word group.

[0074] Using the DTW algorithm, the similarity between the frequency sequence of the spectrogram of the reference word group and the frequency sequence of the spectrogram of the real-time word group is calculated, and the ratio between the matching degree and the similarity between the frequency sequences and the sum of the hyperparameters is used as the user confirmation coefficient.

[0075] Specifically, the user confirmation coefficient satisfies the following relationship:

[0076] ;

[0077] Where, represents the user confirmation coefficient, The time length of the spectrogram representing the reference word group, The time length of the real-time spectrogram representing the real-time word group, The frequency sequence of the spectrogram representing the reference word group, A frequency sequence of a real-time spectrogram representing a real-time word group, Represents the dynamic time normalization algorithm.

[0078] That is, adding In order to avoid the situation where the denominator is zero, the similarity of the speech signal in the time dimension is evaluated by comparing the time length of the reference word group and the real-time word group. The higher the similarity of the time length, the closer the rhythm and speaking speed of the speech signal are; The algorithm compares the frequency sequences of the reference and live word groups to assess the similarity of the speech signals along the frequency dimension. The higher the similarity in the frequency sequences, the closer the pitch and timbre of the speech signals. A higher user confirmation coefficient indicates a greater similarity between the input and reference speech features, and a higher likelihood that they can be confirmed as belonging to the same user.

[0079] In addition, another embodiment also includes:

[0080] Calculate the similarity between the frequency sequence of the reference word group and the frequency sequence of the real-time word group using the cosine similarity algorithm;

[0081] Calculate the absolute difference between the time length of the spectrogram of the reference word group and the time length of the spectrogram of the real-time word group, and normalize the difference based on the ratio of the absolute difference to the reference time length. The difference between 1 and the normalized ratio is used to obtain the degree of match in time length, which is used to measure the similarity in time length between the reference word group and the real-time word group.

[0082] The product of the similarity and matching degree between frequency sequences is used as the user confirmation coefficient.

[0083] Specifically, the user confirmation coefficient satisfies the following relationship:

[0084] ;

[0085] Where, represents the user confirmation coefficient, The time length of the spectrogram representing the reference word group, The time length of the real-time spectrogram representing the real-time word group, represents the cosine similarity algorithm, The frequency sequence of the spectrogram representing the reference word group, A frequency sequence representing a real-time spectrogram of a real-time word group.

[0086] In other words, the calculation of cosine similarity is relatively simple and suitable for large-scale data sets, but it cannot handle nonlinear changes on the time axis, such as expansion or offset of the time axis. When the frequency sequence lengths of the reference word group and the real-time word group are the same, cosine similarity can provide a fast and accurate similarity measurement. For nonlinear changes on the time axis, such as expansion or offset of the time axis, it is suitable for processing speech signals with different speaking speeds. When the frequency sequence lengths of the reference word group and the real-time word group are different, The algorithm can find the optimal matching path through dynamic time warping. When there is obvious expansion or offset between two frequency sequences on the time axis, Algorithms are able to better align and compare sequences.

[0087] S4: Determine whether the real-time input voice segment and the reference voice segment are from the same user based on the user confirmation coefficient, completing the human-computer interaction process of voice recognition.

[0088] In response to the user confirmation coefficient being less than or equal to the preset similarity threshold, it is considered that the input voice segment and the reference voice segment are not from the same user; otherwise, it is considered that the input voice segment and the reference voice segment are from the same user.

[0089] Exemplarily, the preset similarity threshold is 0.5, which can be adjusted according to specific implementation conditions.

[0090] The present invention also provides a human-computer interaction system for speech recognition. Figure 2 As shown, the system includes a processor and a memory. The memory stores computer program instructions. When the computer program instructions are executed by the processor, a human-computer interaction method for speech recognition according to the first aspect of the present invention is implemented. The system also includes other components familiar to users skilled in the art, such as a communication bus and a communication interface. Their configuration and functions are well known in the art and are therefore not described in detail here.

[0091] It should be noted that a person skilled in the art may make a number of modifications and improvements without departing from the scope of the present invention, and these modifications and improvements fall within the scope of protection of the present invention. Therefore, the scope of protection of the patent for this invention shall be based on the appended claims.

Claims

1. A human-computer interaction method for speech recognition, characterized in that: include: Obtain the spectrum information of several historical speech segments of the user, generate a spectrogram using Fourier transform, build the user's speech library, perform word segmentation on each speech segment in the speech library, and obtain a sequence of word units; Using a clustering algorithm to divide the sequence of word units to obtain word groups consisting of multiple consecutive word units, using each word group in the speech database as a reference word group, and using the spectrogram corresponding to each word group as a reference speech segment; The clustering process includes: calculating the segmentation coefficient of each word group based on the mapping function of the upper and lower envelopes of the spectrogram, setting the initial number of clusters, assigning the word units to clusters in order according to the segmentation coefficient, constructing the stopping coefficient for cluster division based on the number of word units and the segmentation coefficient, performing clustering iterations, and outputting the optimal clustering result based on the stopping coefficient; The segmentation coefficient includes: integrating the area between adjacent minimum points of the upper envelope, calculating the signal energy in the area between adjacent minimum points, dividing the integration result by the product of the interval width and the maximum value between adjacent minimum points, normalizing the integration result, summing the normalized results of all adjacent minimum values ​​and dividing by the number of minimum points to obtain the mean of the upper envelope; integrating the area between adjacent maximum points of the lower envelope, calculating the signal energy in the area between adjacent maximum points, dividing the integration result by the product of the interval width and the minimum value between adjacent maximum points, normalizing the integration result, summing the normalized results of all adjacent maximum values ​​and dividing by the number of maximum points to obtain the mean of the lower envelope; averaging the sum of the mean of the upper envelope and the mean of the lower envelope, and subtracting the average from 1 to obtain the segmentation coefficient of each word group; The stopping coefficient includes: calculating the ratio of the number of word units in each word group to the total number of word units in the entire text sentence to obtain the contribution degree of each word group; taking the average value of the product of the contribution degree of all word groups and the division coefficient as the stopping coefficient for the corresponding cluster division; Acquiring spectrum information of the real-time input speech segment and the divided real-time word groups, and in response to the real-time word group being the same as the reference word group, retaining the spectrogram of the real-time word group, and using the similarity between the spectrogram of the real-time word group and the spectrogram of the reference word group as a user confirmation coefficient; The user confirmation coefficient is used to determine whether the real-time input voice segment and the reference voice segment are from the same user, completing the human-computer interaction process of voice recognition.

2. The human-computer interaction method for speech recognition according to claim 1, characterized in that: The word segmentation process is performed on each speech segment in the speech database to obtain a sequence of word units, including: Convert each speech sentence in the speech library to text to obtain the corresponding text sentence. Use the jieba library to divide the text sentence into word units to obtain a sequence of word units, where a text sentence contains several consecutive word units.

3. The human-computer interaction method for speech recognition according to claim 2, characterized in that: The word unit is the basic component of speech or text, and is used to represent the smallest semantic unit of speech signal or text content; the word group is a collection of several consecutive word units, and is used to represent a specific part of speech or text; the text sentence is the complete text content after speech transcription, and is a continuous text.

4. The human-computer interaction method for speech recognition according to claim 1, characterized in that: Outputting the best clustering result based on the stopping coefficient includes: In response to the stop coefficient being greater than the preset stop threshold, or when the stop coefficient reaches the maximum value, the output clustering result is the optimal clustering result, wherein in the optimal clustering result, each cluster corresponds to a word group, and the number of the clusters does not exceed the number of word groups.

5. The human-computer interaction method for speech recognition according to claim 1, characterized in that: The user confirmation coefficient includes: Calculate the absolute difference between the time length of the spectrogram of the reference word group and the time length of the spectrogram of the real-time word group, and normalize the difference based on the ratio of the absolute difference to the reference time length. The difference between 1 and the normalized ratio is used to obtain the degree of match in time length, which is used to measure the similarity in time length between the reference word group and the real-time word group. Using the DTW algorithm, the similarity between the frequency sequence of the spectrogram of the reference word group and the frequency sequence of the spectrogram of the real-time word group is calculated, and the ratio between the matching degree and the similarity between the frequency sequences and the sum of the hyperparameters is used as the user confirmation coefficient.

6. The human-computer interaction method for speech recognition according to claim 1, characterized in that: The user confirmation coefficient also includes: Calculate the similarity between the frequency sequence of the reference word group and the frequency sequence of the real-time word group using the cosine similarity algorithm; Calculate the absolute difference between the time length of the spectrogram of the reference word group and the time length of the spectrogram of the real-time word group, and normalize the difference based on the ratio of the absolute difference to the reference time length. The difference between 1 and the normalized ratio is used to obtain the degree of match in time length, which is used to measure the similarity in time length between the reference word group and the real-time word group. The product of the similarity and matching degree between frequency sequences is used as the user confirmation coefficient.

7. The human-computer interaction method for speech recognition according to claim 1, characterized in that: The determining, based on the user confirmation coefficient, whether the real-time input voice segment and the reference voice segment are from the same user includes: In response to the user confirmation coefficient being less than or equal to the preset similarity threshold, it is considered that the input voice segment and the reference voice segment are not from the same user; otherwise, it is considered that the input voice segment and the reference voice segment are from the same user.

8. A human-computer interaction system for speech recognition, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the human-computer interaction method for speech recognition according to any one of claims 1 to 7 is implemented.

Images