Development of speech recognition system for remote vocal music teaching based on Markov model

With the popularization of smart homes, car audio systems and various speech recognition software, speech recognition systems have gradually entered people’s sights and are favored by most users because of their practicability and accuracy. Cognition is an important interface for human–computer interaction. It will become a research focus in the field of artificial intelligence. It plays an important role in cultivating the basic characteristics of music and cultivating students’ interest in music, and vocal music teaching. Teaching traditional vocal music education to students is in the form of classrooms, such as vocal music, arrangement, and bel canto. The disadvantage is the lack of communication between the classroom and teachers and students. On the other hand, the development of Internet technology provides a new teaching method for traditional vocal music teaching and provides a network infrastructure for building a vocal teaching system platform. Therefore, this article provides a preliminary construction of a remote vocal music education platform by combining vocal music education with Internet technology. The remote audio and video training system is a complex and relatively large project with multiple functions and is to introduce important functions in this system. At the same time, register and log in to the remote voice and video implementation requirements and system functions, respectively, to realize functions such as video training and video-on-demand training.


Introduction
In this article, we will study the recognition of low-altitude and ultra-low-altitude target sound signals in a wireless sensor network detection system, and use software and hardware solutions for speech recognition (Butun et al. 2013). This article analyzes the advantages and disadvantages of hardware and software solutions, and considers their applicability to the system (Salman et al. 2019). At the same time, this article also summarizes the problems in the development process of voice detection circuit and voice chip voice recognition, and evaluates the detection system (Maleh et al. 2015). This article introduces the next step of miniaturization of the detection system and develops a new and improved solution. The solution uses a DSP chip, and the chip supplements the relevant recognition circuit based on the shortcomings of the voice chip solution. The speech recognition algorithm used in this solution proves the speech recognition algorithm used for speech recognition through theory and practice, and proposes a design scheme for the hardware circuit processing of the speech recognition system, which will lay the foundation for us, thereby further improving system performance (Kurzekar et al. 2014). Due to the influence of deep learning in recent years, the acoustic and language models based on deep neural networks are compared with traditional GMM-HMM and n. The Gram model has achieved significant performance improvements (Awwalu et al. 2019). In this case, this article conducts a more in-depth study of the system and at the same time starts with the model structure of the deep neural network (Samek et al. 2016). On the one hand, it optimizes the existing model; on the other hand, it explores the combination of new network structures, and that is, it has the characteristics of speech and language signals (Liu et al. 2017). This combination improves the performance of the deep neural network speech recognition system and the efficiency of training. In order to promote students' learning, the remote vocal music teaching system makes full use of modern information technology and the rapidly developing Internet technology to provide students with more online learning freedom (Sun 2020). We have adopted independent research methods. From a student's point of view, through it, you can learn about system music knowledge that you may be interested in, view learning records, view online dynamic learning information, and view the teacher's recommendation process to make it easier for teachers to implement online education to understand the information (Lu 2014). The task of the class is changed. It provides functions such as public lectures, learning exchanges, and information management. From the teacher's point of view, they can manage speeches, manage students, publish and view dynamic information about online learning, manage examination questions and answers, view student scores, and open (or close) the website function in student communication (Kallio and Heimonen 2019).
In the design process, we chose the Internet-based Java language, because the music distance learning system is an Internet-based system. It uses MVC architecture and SSH framework technology to implement the software to improve the efficiency of Java usage.

Related work
The literature introduces the advantages and disadvantages of software and hardware solutions for recognition, and at this stage, a solution suitable for the system's speech recognition software and hardware is discovered (Hodgson and Coiera 2016). We use an electret microphone as a sensor to detect the target sound signal in the air, and use a high-sensitivity sound sensor circuit to accurately detect the sound signal. This article summarizes the voice chip recognition method and the problems in the development process, and conducts experiments to determine the voice recognition module and detection system (Principi et al. 2015). The literature introduces a new and improved method in which DSP chips are used and identified according to appropriate schemes. By comparing the acoustic signal signature mechanism with the language creation mechanism, the voice chip solution was successfully used to achieve low target signals and ultra-zero target signals (Al Atik and Abrahamson 2010). Speech recognition algorithms have been proven in theory and practice. The literature introduces the basic theory of speech, starts speech attribute extraction, and combines the principle of deep learning to construct speech attribute extraction, and uses the extracted speech attributes to make a speech recognizer (Devi and Suganya 2016). Compared with the built-in speech recognition, it greatly improves the recognition function of consonants and vowels and the recognition rate of consonants and vowels, and reduces the error rate of word recognition (Healy et al. 2013). Multiconstraint and nonlinear configuration problems are used to illustrate the grid allocation problem, and the relative difference is used to find the best solution for the best transmission radius of the nodes in the network. The real-time history of the network, regular node information and healthier feedback (Li et al. 2016).
3 Wireless sensor network and speech recognition system

Energy consumption control algorithm model
In this section, the network model used in this algorithm introduces the node model and the energy model, respectively. Since the research of this paper mainly focuses on the energy imbalance of wireless sensor networks, this paper is based on an ideal channel model, which is helpful for problem analysis. Ensure that there is no channel conflict and that both networks are within range, and internal nodes can communicate with each other. Since N sensor nodes are randomly and evenly distributed in a circular surveillance area with a radius of R, the radius of the surveillance area can be expressed as: Among them, m is the number of concentric circles other than the outermost ring, w is the width of the ring, and c is the width of the outermost ring. The width of the ring will not change in the simulation experiment.
The whole area is divided into S-shaped rings, which can be expressed as: The node density q of the network can be expressed as: The energy consumption of a sensor node can be roughly divided into transmission energy consumption E t generated by data transmission, receiving energy consumption E r generated by data reception, and energy consumption E s when sensing data. The main part of the energy consumption of E t and E r is generated by node i, and the transmission energy consumption E t for sending kbit data to node j can be defined as follows: The received energy consumption E r can be defined as: The sensor nodes are randomly and evenly distributed in the monitoring area. The number of i-ring nodes can be expressed as follows by the network node density and the area occupied by the i-ring: The data transmission model can be defined as: Re k is related to the sending radius of the node and the width of the ring.
Therefore, the parameter k can be expressed as: All rings must complete the transmission task in order to achieve the lowest energy consumption that can be defined as: A set of restrictions can be formally expressed as: Can be defined as: The strictest function G(X) can be expressed as: When X is a discrete variable, the relative difference value can be expressed as: Through the above formula, it can be expressed by formula (16).
Energy balance factor (EBF). EBF is used to measure the energy balance of each ring. It is displayed as the standard deviation of the remaining energy for each ring.
Packet loss rate (PLR). The packet loss rate is an important indicator to measure the effectiveness of the algorithm. If the packet loss rate of this algorithm is very low, performance will be affected. In this article, the packet loss rate can be defined as: Node mortality (NDR). Use node mortality to evaluate the balance of energy consumption in the algorithm. At the end of the network life cycle, the node mortality curve means that the energy consumption of steep nodes is close to the energy consumption of network nodes and is in a balanced state. The sensor is an approximation. The more the curve is shaken, the better the balance of energy consumption. The death rate of the nodes in the i-ring is defined as: First, we compare the energy consumption and life cycle of the nodes of the three algorithms in the 0 ring. Compared with the short path and DTA algorithm, PLFC extends the network life cycle by 25.2% and 22.3%, as shown in Fig. 1. This algorithm effectively prolongs the life cycle of the network.
In Fig. 2, the energy consumption balance in the loop of the DTA algorithm is better than the other two algorithms. The short path algorithm has the lowest energy consumption balance, and PLFC is somewhere in between. However, since the energy consumption in the DTA and short path algorithm loop is faster than that of PLFC, the inflection point of the DTA and short path curve is shown earlier than the inflection point of PLFC.
The advantages of the PLFC algorithm in Fig. 2 are not yet clear, but the performance analysis combined with the packet loss rate curve provided in Fig. 3 shows that PLFC is superior to DTA. The PLFC algorithm considers the redistribution of the load, so compared with the short-path algorithm, it is recommended to balance the energy consumption of the PLFC. When choosing a path, PLFC will not only choose adjacent nodes close to the base station, but also consider energy consumption. Therefore, the packet loss rate is lower than the short path algorithm. Next, we analyze DTA. When DTA starts to operate the algorithm, its packet loss rate is very high. This is because the path-finding algorithm needs to be initialized, and each source node must choose a path to find the target node. This process introduces additional energy consumption and optimizes the transmission radius for the nodes on the path. The DTA algorithm has a lower packet loss rate than the short-path algorithm, because in the later stage of the network life cycle, the remaining energy of the nodes in the 0 ring is higher than the energy of the short-path algorithm. Therefore, the illusion of energy balance in DTA is not due to load redistribution, but due to high packet loss rate. The DTA life cycle of the route search algorithm is shorter than that of PLFC.

Speech recognition system
The main function of speech recognition is to convert speech signals into text information, which is mainly composed of acoustic characteristics, speech models, acoustic models and codecs. The learning and recognition process is the learning of extracting acoustic characteristics from the audio data of the original waveform to obtain the network acoustic model formed with the utterance dictionary and language model. New speech features, acoustic models and recognition functions provide Viterbi decoding results.
In all hidden Markov models, large-scale vocabulary continuous speech recognition systems are statistically trained. It will output the word sequence based on the maximum posterior algorithm. The mathematical expression is as follows: Equation (20) can be changed to: Kepstrum coefficient is an important audio characteristic parameter realized based on the isomorphic processing method. The processing method is as follows.
HMM can be described as five parameters: It is difficult to explain the characteristic distribution of speech signals with a simple Gaussian probability density function. In practical applications, the mixed Gaussian model is usually used to match the audio signal, and the output probability is mainly expressed by the mixed Gaussian function: Statistical language models use probabilities instead of judgments based solely on grammatical rules to express the likelihood of word sequences appearing in the language environment. When the word sequence W = {W 1 , W 2 …, W n }, the value of the probability of occurrence can be expressed as follows.
Equation (26) becomes: 4 Design and practical application of longdistance vocal music teaching system

System requirement analysis
The music education system aims to provide students and teachers with an intensive learning platform, so that teachers can display some educational resources and make some learning plans. Students can devote themselves to overall learning, download and browse online learning resources, and complete various learning tasks. The above-mentioned characteristics are essential elements of the general learning support service system. However, in this learning mode, students may suffer from learning fatigue due to lack of certain inducements. Without relevant supervision and surveillance, people will not be enthusiastic about online learning. The learning effect is not ideal. Therefore, the music education system studied in this paper introduces functions such as online music learning and feature learning, and provides learners with a comprehensive music learning platform through regular learning, homework, and information notification. Practice on the Internet and improve the learning effect by changing the previous model. The system can also be applied to other specialized learning systems. We can apply and promote the function and dynamic model of the teaching system learned in the paper. It can be used and improved in the online learning system, similar to the student learning and hands-on mode. Mainly through independent study and learning interest to enable students to better acquire knowledge. Through analysis, it can be known that the system at this stage also needs to include the following aspects: (1) Basic information management requirements The music education system is aimed at students and teachers. In addition to these two users, there is also the role of system administrator. System roles require students to complete personal registration on the system. Registration information includes basic personal information, such as after registering name, gender, date of birth, race, political views, profession, home address, health status, accommodation, and contact information; students will not be able to complete learning from the platform immediately, but waiting for review and certification of student information.
In addition, in the system's music teaching system, the personal information of these teachers always exists in the form of Excel files, so there is no need to enter the teacher's information. It must support the import of Excel files through the system and call the basic information of the teacher in the file reading format. After completing the batch import of teacher information, the teacher will use his ID number or mobile phone to complete the system login. The login process can be completed using the mobile application or the computer desktop. Therefore, teacher management must complete the input, modification and deletion of teacher information, and provide the function of importing batch data.
(2) Student music assignment management needs In the curriculum system, in addition to the normal study of the students, the teacher also needs to assign some homework. Music homework is the foundation of student learning and teacher teaching. Teachers must complete the correction of music homework through this system. Music homework is implemented through language or file attachment format, allowing teachers to keep their music homework and download or delete their music homework. After the teacher uploads the music homework, the students can search and download the system homework.
(3) Music practice management needs Music is a practical subject. In addition to a good grasp of the basic theories of learning music, it is also necessary to master music-related content through music work training and practice through mobile phones. The system is famous because it emits sounds through the mobile phone to reflect the effects of the exercises. It is necessary to learn famous works to improve professional standards. At the same time, it is necessary to provide data on various works.
(4) Online classroom management needs Learning in the classroom is an important part of the system. Learners can order learning resources online, or ask teachers a variety of research questions.

(5) Information notification management requirements
In the music education system, student participation in learning is a process. Teachers can complete the release of educational information and notification dynamics through the platform and push them to each student's mobile phone. Information notifications usually include educational information, daily notifications, and information notifications. The system needs to provide students with educational information displayed in news and dynamic texts, which can be inquired in mobile applications. Educational information includes browsing education information, publishing education information, deleting education information, etc. Teachers can post educational information, but they cannot be changed or deleted after posting.

Basic information management function model
Through demand analysis, we can know that the end users of the system through the school and the platform include students, teachers and system administrators. Generally, when a student starts a music class, the student must complete registration on the platform via a mobile device. After inputting, the teacher will verify the background of the student information and submit it. If it is approved, the student registration information will be valid, and it will be possible to log in through the client.
User information management includes user information management, user information display and user statistical information. Basic information management is a use case diagram.
(1) User information maintenance: Information is the foundation of the system. Detailed user information is always available throughout the system. User information management includes operation, modification and deletion of new functions. If a new user needs to pay, the user's administrator must enter the user's detailed information. If the user information has been changed or an error message appears, it needs to be changed and completed by correcting it; when the user logs out, the user information must be deleted.
(2) User information query: The user can complete the user's query through keywords or a combination of multiple conditions. (3) User information statistics: statistical information provided for each user through educational background and classification, such as statistical information about the number of users with educational background and student age, and operating users can set their own statistical conditions.

The functional model of student music homework management
Music homework is the basis of teaching for students and teachers. Therefore, in order to maintain the system's music homework, we need to manage music homework in the system. Music assignments are videos, audios, and other related materials commonly used in online learning systems. It is an online learning system, so music assignments, text descriptions, and attachments are expressed in formats.
The management of musical works includes tasks related to musical operation query, musical works uploading and musical works maintenance. In this system, the use of music homework is one of the indicators of student performance evaluation. There are two types of music homework: one is music homework for independent learning, and the other is a resource for learning performance. Students can also search for target courses through course classification, and directly search for resources to achieve their goals. Users can use management tools to design their own personalized process and can use learning tools to complete the learning process. Music assignments include music assignments, publishers, release dates, save passes, etc. And the teacher will introduce the music homework so that students can complete the music homework easily.

Music practice management function model
Students use their mobile phones to practice music, and use their mobile phones to emit sounds to reflect the results of the exercises, and at the same time reflect the process of practice effects such as famous works.

Online classroom management function model
Learning in the classroom is an important part of the system. Learners can order learning resources online, or ask teachers a variety of research questions. Classroom learning includes classroom learning, learning thinking, online questions, resource exploration, etc.
(1) Classroom learning is used to select and play music assignments online, and the system will automatically record the start time of learning.
(2) Learning reflection is used to summarize and reflect the problems related to students in the learning process.
(3) For online questions, the teacher answers the learning questions raised by students online. (4) Resource browsing is to query the status of music works by entering keywords.

Information notification management function model
In the music education system, teachers can complete the release of educational information and notification mechanism through this platform, and push it to each student's mobile phone through push. Information notifications include educational information and daily notifications.
(1) Educational information: In the process of music education, there is a lot of information about music education, such as application policies for music majors and national development. Through educational information, students can use the latest music majors and overall educational information. Educational information is released by teachers and system administrators.
(2) Daily notice: Teachers can issue notices to inform students of some important educational matters.

System structure design
The functions and non-functions of the system are analyzed in detail above. The important functions that the system needs to achieve are basic information management, student music homework management, music practice management, and online classroom management information. It is part of the system design. According to the functional analysis of the system, complete the system design of the entire module and functional modules. The overall design is an overview design of the system, which is related to the system architecture design and data, business logic and functions. Complete the system architecture design from an independent perspective. The physical structure design shows the connection structure between the server and each system in the network topology design of the system. Class diagrams, sequence diagrams, etc., are used in detailed system design. Through detailed explanation of each function of the system, the database design of the system is finally completed.
The system uses the JavaEE framework structure, and the client uses Android. In order to provide maintainability and scalability of the system, the specific structure is shown in Fig. 5 The audio-visual teaching system is mainly composed of server components and client components. Clients are mainly teachers and students' computers. The client mainly collects multimedia data for the user's video and audio, while the server mainly stores data for multimedia broadcasting, as shown in Fig. 6.

System data sheet design
Through system analysis and system overview design, we have a detailed understanding of the functions realized by the system. Database support system requires a complete system database design, and the physical design of the database is the core structure.
The database table design is a part of the database physical design. This part uses the database table design to implement the database physical design by describing the structure of the data table. In this section, we will choose some tables to explain its physical structure.
(1) User information table The user information table is used to store the system terminal user information and the system background desktop system. Table 1 lists the fields, types, and physical storage names in the table.
(2) Information sheet Table 2 lists the fields, types, and physical storage names in the table. This information is used to store detailed information about the information.
(3) Attachment information table   Table 3 lists the fields, types, and physical storage names in this table. The attachment table is used to store attachments of information and documents related to notifications. The attachment information of the system is stored in a file format.  (4) Daily notification information table   Table 4 lists the fields, types, and physical storage names in this table. The daily notification information table is used to store detailed information about daily notifications.
(5) Job announcement information form Table 5 lists the fields, types, and physical storage names in the table. The job list table is used to store related job lists in the system.   (20) Foreign key, cannot be empty WDBH (1) User information management: User information includes user name, user number, user address, etc. This basic information is the basis of the system, where you can find detailed user information. System-wide maintenance includes adding, modifying and deleting tasks. If a new user is included, the user's administrator must enter the user's details. If the user information is changed or an error occurs, the user information needs to be changed, the change is completed, and the user information must be deleted when the user logs out.
(2) Query user information: input user keywords or combine multiple conditions to complete user query. A simple query is just a keyword that completes the user's query, while a compound query is two or more, which is completed by combining words. (3) User information statistics: Based on the statistical information of each user based on learning ability and classification, users can set their own statistical conditions.

Conclusion
In order to better research on the construction of distance learning systems, we will first analyze and study the theories of distance learning professors and then find suitable distance learning teaching theories. Secondly, we analyzed and studied the technology required for the design of the music remote system, and determined the design tool of the software system according to the needs of the music system. In the design process, we chose the Internet-based Java language, because the music distance learning system is an Internet-based system. It uses MVC architecture and SSH framework technology to implement the software to improve the efficiency of Java usage. Applying the SSH framework technology to the music remote system allows us to layer the system and also helps us expand and transplant programs. The database uses open-source MySQL and uses Toad for MySQL visual database operation software to accelerate database development. The system passed the test operation, and the feedback effect of the system is good. The page design style and layout structure are reasonable, easy to use, and clear and smooth execution speed provides students with a good experience. These functions of online learning, online testing and viewing the learning process are both practical and easy to use, allowing students to freely choose time and place for independent learning. Interactive online question and answer provides a new way for students to communicate with teachers. The system can easily realize the accuracy of student management, and through increasing teaching resources and information, the system has important practical significance in education and learning.
Funding Data availability Data will be made available on request.

Declarations
Conflict of interest The authors declare that they have no conflict of interest.
Ethical approval This article does not contain any studies with human participants performed by any of the authors.