Application of Android Voice Assistant based on Parallel Storage in Multimedia English Teaching Platform

doi:10.21203/rs.3.rs-2756533/v1

In recent years, supercomputing technology has become more and more important in scientific research and economic construction, and its application fields have become more and more extensive. Supercomputers and data storage are entering the E-class era, the computing power of supercomputer systems is rapidly increasing, and the amount of data is also growing rapidly. The contradictory relationship between parallel storage and computing has become more and more obvious, which poses a major challenge to the processing capabilities of supercomputer systems. Android speech recognition technology has great theoretical value and broad application prospects. In recent years, it has received widespread attention and achieved great success. The research of artificial neural networks (especially deep neural networks) has greatly improved the accuracy and speed of speech recognition, and has shifted speech recognition technology from laboratories to commercial systems, gradually affecting and improving people's work and life. The network teaching platform based on computer technology is not only a tool that can help teachers share knowledge, but also a tool that encourages students to learn independently, emotional stimulation tools, and a rich teaching environment. These tools have been organically integrated into the classroom environment and become traditional teaching methods have brought fundamental changes. The multimedia English course combines a variety of functions, such as text, graphics, images, sound, animation, and video. It has multimedia presentations, personalized teaching methods, diversified teaching methods and autonomous learning environments.

Parallel storage

Voice assistant

Multimedia

English teaching platform

As more and more information becomes electronic information, storage technology becomes more and more important [1]. In particular, the rapid growth of computer network applications has greatly increased the demand for information storage products and the security and reliability requirements of information storage technology [2]. According to a company study of 800 internet-connected microcomputer users, every hard drive failure will result in more than 5 days of ineffective working days. In a typical commercial application, it takes an average of 3.5 months to recover 1,000MB of data, and the cost is US$95,000. In recent years, storage technology has developed rapidly [3]. Various new products and technologies emerge in an endless stream, but in general, they have a pyramid structure, where the spire is the CPU. The closer to the CPU, the higher the memory speed per megabyte and the smaller the storage capacity. Conversely, the lower the storage speed, the lower the storage cost per megabyte and the greater the capacity. Including the development history of speech recognition technology, the development status at home and abroad, as well as the content and research purpose of this article. In addition, the basic structure and principle of the artificial neural network algorithm are described in detail [4]. Then, according to the software development ideas, define the software requirements and functional design, and develop a complete software solution for the voice assistant, including mobile phone recording, file storage and Android Bluetooth communication, client-server interaction, database query and modification, and Phone and text messaging operations [5]. Finally, all the functions of the voice assistant software are implemented on the Android platform and voice recognition algorithms [6]. The storage system is the software and hardware system in the supercomputer system, responsible for storing data and supporting data input and output applications [7]. The efficiency of accessing storage system data is a key factor affecting the actual performance of supercomputing system applications [8]. In the era of big data, many supercomputing applications have evolved into I/O-intensive or I/O-intensive applications that require large amounts of storage, processing power, and I/O bandwidth. Data-intensive applications, thereby improving storage I/O performance [9]. Voice recognition technology is a hot spot in current scientific research, and voice assistants are also a popular trend in the development of current mobile clients. In this paper, artificial neural network is used as the research background to systematically study the speech recognition process. According to the theoretical basis and principles of neural network training, the focus is on the speech recognition of isolated command words based on artificial neural network, and the corresponding algorithm is completed [10].

The literature [11] mentions that the hard disk is the most popular storage medium for most users, and the hard disk is an indispensable device in all modern personal computers, and due to its portability and versatility, floppy disks are still the first medium for computer users to exchange data. External storage has the fastest access speed, so it is more suitable for storing programs and files that require frequent and fast access. For individual users, hard drives are the most important storage devices. The computer's operating system, applications, and important data are stored on the hard drive. Therefore, the capacity of the hard drive should be as large as possible. The literature [12] shows that speech recognition is a technology in which machines convert human speech signals into corresponding texts or commands through recognition and understanding. Or, it can open an application on a mobile phone and develop a security system that uses voice fingerprint recognition and user identity verification. The literature [13] shows that the rapid development of information technology and the Internet has contributed to the popularization and use of multimedia teaching platforms. In the five aspects of English listening, speaking, reading, writing and translation, multimedia and other auxiliary learning methods play an important role. The literature [14] shows that under the background of information technology and technology, in order to ensure the smooth progress of various reform projects in China, this article needs to make full use of modern educational technology in college English teaching and autonomous learning to make greater progress. The systematic English teaching system makes the English learning mode more personalized, and gradually improves the English teaching level through the multimedia network [15]. College English has entered a new era of teaching. It uses computer networks for teaching and expands the teaching experience from a limited classroom to an unlimited network.

3.1. Parallel storage system structure

Distributed shared parallel storage can be expanded on a large scale and can provide continuous high-availability processing bandwidth. At the same time, the architecture and file system are more open, which is conducive to user management and secondary development. As a result, most modern supercomputing centers use a distributed shared parallel storage architecture. The block diagram is shown in Fig. 1.

Considering the structural characteristics of massively parallel storage, an ideal parallel file system model was first created. The ideal file system modeling is largely based on the architecture, as shown in Table 1:

Table 1

The meaning of each parameter
Parameter	Meaning	Parameter	Meaning
Ci	A computing node	Wn	Network bandwidth
C	Compute node collection	ΔTm	Delay in processing an I0 request by the metadata server
M	Metadata server	Ws	Maximum I0 bandwidth of the data server
Sj	A data server	Vi	Calculate the file size requested by a certain I0 of the node
L	Number of compute nodes	K	Number of data servers
S	Data server collection	O_j	Correlation conflicts caused by multiple IO requests
	The access size is V, the number of IO requests generated	D	The size of the data requested by each I0
PJ	I0 performance when there are j conflicts (the larger the I0 performance, the worse)	T_m	Access of size V, time overhead required by the metadata server
Tn	Data delay through the network	T_s	Data server read and write time overhead
ηc	Cache hit rate of data in computing node	η₀	The cache hit rate of the data on the data service
Dc	Delay overhead for a client i to access data	W_c	Bandwidth of a client i to access data

Use an ideal mathematical model to describe the process of the client accessing the data server in the parallel storage system, as shown in Fig. 2.

If L computing nodes access the metadata server at the same time, the data server waiting time of the computing node C_i can be expressed as formula (1):

$${T}_{M}\left({V}_{i}\right)=\rho \left(L\right)\phi \left(\sum {V}_{i}\right)\varDelta {t}_{m}$$

1

And assuming that there is not much difference in the cache hit rate, formula (2) and formula (3) can be obtained:

$${\eta }_{{C}_{I}}={\eta }_{{C}_{J}}={\eta }_{{C}_{O}}$$

2

$${\eta }_{{O}_{I}}={\eta }_{{O}_{J}}={\eta }_{{O}_{O}}$$

3

The parameters of the hard disk performance degradation caused by the data access conflict on the medium arriving at the data server, such as formula (4):

$$Q=L\times \left(1-{\eta }_{C}\right)\times \left(1-{\eta }_{0}\right)$$

4

Since the original data server still has a large amount of post-discovery locking overhead, the busy time is very long, so the metadata conflict overhead is as follows: formula (5), formula (6), formula (7), formula (8), formula (9):

$$Q=L\times \left(1-{\eta }_{C}\right)$$

5

$${T}_{N}=\frac{{V}_{i}\times \left(1-{\eta }_{{C}_{i}}\right)}{{W}_{N}}$$

7

$${T}_{X}=\frac{{V}_{i}\times \left(1-{\eta }_{{C}_{i}}\right)\left(1-{\eta }_{0}\right)}{{K}_{i}{W}_{S}}$$

8

$${D}_{Ci}={V}_{i}\left(1-{\eta }_{Ci}\right)\left(\frac{\rho \varDelta {t}_{m}}{D}+\frac{1}{{W}_{N}}+\frac{1-{\eta }_{oi}}{{K}_{i}{W}_{S}}\right)$$

9

Since the network is non-blocking, the average network delay is formula (10):

$$\stackrel{\_\_\_}{{T}_{N}}=\frac{\sum \left({V}_{i}\times \left(1-{\eta }_{Ci}\right)\right)}{L\times {W}_{N}}=\left(1-{\eta }_{C}\right)\times \frac{\stackrel{\_\_}{V}}{{W}_{N}}$$

10

After the storage system was built, the storage system was tested. The configuration of the system structure is shown in Table 2.

Table 2

Test environment
Equipment name	Quantity	RAM	Description	Storage
IO Service Acceleration Node (ION)	512	8X16G	Dual-channel high-energy Intel Xeon CPUE5-2692V2, 12 cores, 2.2GHz; 1 self-developed high-speed interconnection network interface, 1 IB interface (the former is connected to the computing node, the latter is connected to the storage network) 2 ITB capacity PCIE solid state memory cards .	2T solid state storage
Client (computing node)	19840	12X16G	Dual-channel high-performance Intel Xeon CPU E5-2692V2, 12 cores, 2.2GHz; 1 self-developed high-speed interconnection network interface, and an IB interface.	0
Self-developed network	The measured bandwidth of a single link is 12.6GBps.

The client starts reading and writing to an ION, and the corresponding number of computing nodes ranges from 1 to 128, with one process per node. The test results are shown in Fig. 3.

Figure 3 shows that the read and write bandwidth shows a linear growth trend. Therefore, if the number of tested clients exceeds 32, the throughput begins to decrease. By default, every 64 clients are counted. If the user program needs more I/O bandwidth, the administrator can increase the I/O by changing the HVN. Due to the insufficient number of computing nodes in Test 64, some computing nodes have run two tests. The test results are shown in Fig. 4:

The test environment is shown in Table 3.

Table 3

Test environment
Equipment name	Quantity	RAM	Description	Storage
IO Service Acceleration Node (ION)	512	8X16G	Dual-channel high-energy Intel Xeon CPU E5-2692V2, 12 cores, 2.2GHz; 1 self-developed high-speed interconnection network interface, and an IB interface. (The former is connected to the computing node, and the latter is connected to the storage network) 2 ITB capacity PCIE solid state storage cards.	2T solid state storage
Client (computing node)	19840	12X16G	Dual-channel high-performance Intel Xeon CPU E5-2692V2, 12 cores, 2.2GHz; 1 self-developed high-speed interconnection network interface.	0
Storage server	128	12X16G	Dual-channel high-performance Intel Xeon CPU E5-2692V2, 12 cores, 2.2GHz; 1 self-developed high-speed interconnection network interface, and an IB interface.	A total of 64 IB-SANs, each with a storage capacity of 64*4T
Self-developed network	The measured bandwidth of a single link is 12.6GBps.

3.2. Technical basis of Android voice assistant

The main process of the dual-threshold endpoint detection algorithm is: after reaching a certain intensity of short-term average energy and crossing the short-term zero frequency, first set a low threshold for the weaker signal, and then set a high threshold for the signal. When the average energy and the zero-crossing frequency both return below the lower threshold, the end position is noted, and the energy or zero-crossing speed of the subsequent frame signal is also continuously measured. If the values of the two parameters increase and exceed the upper threshold, it can be determined that the effective voice signal has not ended, and the currently marked end point has been deleted, but the two test indicators have not been exceeded. Since there can be multiple characters in the control word, the audio signal repeats the process of starting, ending, and restarting. It is necessary to find the start point of the first mark and the end point of the last mark to determine the exact length of the control word. The block diagram of the double threshold detection algorithm is shown in Fig. 5.

Linear activation function, such as formula (11):

$${y}_{k}=f\left({v}_{k}\right)={K}_{vk}$$

11

Threshold logic activation function, such as formula (12):

$${y}_{k}=f\left({v}_{k}\right)=\left\{\begin{array}{c}1\\ 0\end{array}\right.\begin{array}{c}{v}_{k}\ge {\theta }_{k}\\ {v}_{k}<{\theta }_{k}\end{array}$$

12

S-type logic activation function, such as formula (13)

$${y}_{k}=f\left({v}_{k}\right)=\left\{\begin{array}{c}1\\ 0\end{array}\right.\begin{array}{c}{v}_{k}\ge {\theta }_{k}\\ {v}_{k}<{\theta }_{k}\end{array}$$

13

Hyperbolic tangent activation function, such as formula (14):

$${y}_{k}=f\left({v}_{k}\right)=th\left({u}_{k}\right)$$

14

The output data of the neurons in the output layer, as in formula (15):

$${y}_{j}={f}_{2}\left({\sum }_{k=1}^{q}{w}_{jkZk}\right)$$

15

The error E_p of the p-th sample will be Eq. (16):

$${E}_{p}=\frac{1}{2}{\sum }_{j=1}^{m}{\left({t}_{pj}-{y}_{pj}\right)}^{2}$$

16

For all examples, the global error is formula (17):

$$E=\frac{1}{2}{\sum }_{p=1}^{p}{\sum }_{j=1}^{m}\eta \left({t}_{pj}-{y}_{pj}\right)={\sum }_{p=1}^{p}{E}_{p}$$

17

Define the error signal, such as formula (18):

$${\delta }_{xj}=-\frac{\partial {E}_{p}}{\partial {S}_{j}}=-\frac{\partial {E}_{p}}{\partial {y}_{j}}\cdot \frac{\partial {y}_{j}}{\partial {S}_{j}}$$

18

4.1. Multimedia English teaching platform design

According to the principles of teaching constructivism, the learning process is no longer a one-way transfer of knowledge. Studying English at university must be supported by multimedia network technology in order to create a real English learning environment for students so that they can study English at university. In a real English learning environment, students are reorganizing and reforming their knowledge and experience, and actively constructing meanings to acquire new knowledge. Based on the teaching theory of humanism, the design of education platform should be individualized.

English listening courses are usually held in the language area, and the education of listening materials is carried out under the guidance of the teacher. No language media can convey specific sounds, images or texts. If you really want to improve the students' English level, you need to create a real learning environment. "With the deepening of information theory and cognitive learning theory, the process of listening comprehension is regarded as a positive information processing process." The constructivist theory in the classroom believes that the acquisition of students' language skills must go through a process of meaning construction. In a realistic language learning environment, students use two-way visual and auditory stimuli to construct the meaning of knowledge. Multimedia technology also enables teachers to use computers to monitor the progress of students and ask questions. Various auxiliary learning platforms such as video, audio and animation also play an important role in college English teaching and offline learning.

Oral expression is the first priority in English learning, and it is also an important indicator for assessing English proficiency. Of course, the teaching of oral English is inextricably linked with the teaching of listening, reading, writing and translation. These five aspects of learning are complementary to each other. Through listening, reading, writing and translation teaching to improve oral English, on the contrary, improving oral English will help listening, reading, writing and translation. Learn about the relevant Western cultural background through the Internet and other resources, and prepare for classroom activities hosted by classroom teachers. Teachers arrange classroom performances for students so that students can practice their oral skills during class and entertain them in a fun-filled way. In this way, it creates a real English practice environment for students, gets rid of the dumb English of traditional test-oriented education, and cultivates students' practical application ability of spoken English.

Scaffolding teaching theory means that teachers create a theoretical framework for students to make it easier for students to understand knowledge and gradually transfer learning tasks to students. When the students have completely absorbed the knowledge they have learned, the teacher removes the brackets. This article focuses on the interest in learning, creating a framework for translating text, creating a framework for students to enter the learning environment, and step-by-step instructions for completing translation exercises. In addition, some translation software is becoming more and more complex, such as Google translation software, Lingoes, etc. Students use this translation software to practice their own translation, whether it is vocabulary, phrases or long and complex sentences. In this translation software, the correct translation table will be found, eliminating tedious dictionary searches and making translation learning more efficient.

4.2. Realization of multimedia English teaching platform

If you can listen to English radio or English songs for half an hour every day, which is an indispensable part of life, then radio stations such as BBC, CNN, NPR, VOA will provide popular songs, news and information for this article. This article can also search for original English songs on Kuwo Music or Baidu Music. Even if this article is not very familiar with the content, long-term understanding of the English environment will help this article to correctly understand pronunciation and intonation. However, if they rely on knowledge and vision to listen, they can acquire 65% of the knowledge. From this perspective, it is very important to familiarize the audience with materials such as movies, TV, and online videos. The use of multimedia classrooms and English movies as a teaching platform for listening teaching is mainly due to the fact that English movies contain text, images, sounds, background culture and other elements. Their teaching effect is not superior to that of any other movies. It expands students’ knowledge and vocabulary.

The impression of content and the content of short-term memory are enhanced, so that students have enough time to reflect. Combined with the dual display of Chinese and English subtitles, this makes it easier for students to understand, making oral learning easy and effective, and the software and reading functions enable this article to perform follow-up exercises in a quiet and tense situation. A free environment allows this article to improve spoken English quickly and easily. In addition to follow-up learning software, you can also use online chat software to practice conversational English. Omegle is an anonymous English speed dating chat room created by 18-year-old Leif K-Brooks from Florida, USA. After entering the site, click "Start Chat", and the server will automatically randomly select Internet users to chat. The main advantage is that you can reliably learn spoken English here. Due to the limited learning environment in some schools, especially for those schools that do not speak English, there are no professional foreign teachers to attend classes. In this case, the teacher can show the original English movie or TV show fragments to the students in the classroom. By watching excellent movies and TV works, students can understand the correct pronunciation, intonation and grammar skills of English, as well as some everyday language. Short sentences represent different meanings in different situations, absorb knowledge of Western culture in a real language environment, and help students pronounce them correctly to improve their oral English skills. Regarding teaching students how to pronounce correctly, teachers can not only play movies, but also make or download Flash animations. The teacher uses the Flash program to view the changes in mouth shape, tongue position and vocal cords due to correct English pronunciation. The state of vibration, etc. These changes in pronunciation are indeed presented to the students in the form of animation, accompanied by the teacher’s demonstration and explanation, which not only reduces the number of learning tasks for the teacher, but also improves the learning effect. More than one. Unilateral explanation.

Ace Reader Pro is an English reading software that enables students to conduct self-tests, improve reading speed, correct reading errors and provide visual movement training in the form of games. When using Ace Reader Pro for the first time, the system will automatically record the user's speech rate, intonation and pronunciation. When reused, the system will automatically accelerate to twice its original speed. Students will learn fast reading and gradually increase their reading speed in the area. If the Ace Reader Pro software is installed on the mobile training terminal, independent training can be carried out at any time.

When developing multimedia listening tools for English, teachers can make full use of multimedia technology to create realistic life scenes for students in the classroom. Students use visual and auditory bidirectional stimuli to understand the meaning of knowledge in this real language learning environment. During the construction process, the teacher tracks the student’s learning progress and learning effects through multimedia teaching resources. Students can also use videos, audios, animations and other supplementary teaching materials to learn English on their own. The use of solid-state drives in storage has achieved a breakthrough in performance, which can be achieved by storage per unit area, and its environmental protection and energy-saving features have also deeply attracted the attention of users. In the future, as the market develops, prices will continue to fall. The continuous expansion of market awareness will completely change the storage status of this article.

Compliance with Ethical Standards

Conflict of interest

The authors declare that they have no conflict of interests

Ethical approval

This article does not contain any studies with human participants performed by any of the authors.

Data Availability

Data will be made available on request.

X. Tan, Q. Li, and H. Wang, “Advances and trends of energy storage technology in microgrid,” International Journal of Electrical Power & Energy Systems, vol. 44, no. 1, pp. 179-191, 2013.
T. Lehtola, and A. Zahedi, “Solar energy and wind power supply supported by storage technology: A review,” Sustainable Energy Technologies and Assessments, vol. 35, pp. 25-31, 2019.
K. Liu, and L. J. Dong, “Research on cloud data storage technology and its architecture implementation,” Procedia Engineering, vol. 29, pp. 133-137, 2012.
Y. C. Wu, and J. W. Feng, “Development and application of artificial neural network,” Wireless Personal Communications, vol. 102, no. 2, pp. 1645-1656, 2018.
A. Poushneh, “Humanizing voice assistant: The impact of voice assistant personality on consumers’ attitudes and behaviors,” Journal of Retailing and Consumer Services, vol. 58, p. 102283, 2021.
S. M. C. Loureiro, A. Japutra, S. Molinillo, et al., “Stand by me: Analyzing the tourist–intelligent voice assistant relationship quality,” International Journal of Contemporary Hospitality Management, vol. 33, no. 11, pp. 3840-3859, 2021.
J. Ousterhout, A. Gopalan, A. Gupta, et al., “The RAMCloud storage system,” ACM Transactions on Computer Systems (TOCS), vol. 33, no. 3, pp. 1-55, 2015.
W. Gentzsch, D. Girou, A. Kennedy, et al., “Deisa—distributed european infrastructure for supercomputing applications,” Journal of grid computing, vol. 9, no. 2, pp. 259-277, 2011.
C. P. Chen, and C. Y. Zhang, “Data-intensive applications, challenges, techniques and technologies: A survey on Big Data,” Information sciences, vol. 275, pp. 314-347, 2014.
S. K. Gaikwad, B. W. Gawali, and P. Yannawar, “A review on speech recognition technique,” International Journal of Computer Applications, vol. 10, no. 3, pp. 16-24, 2010.
B. Sprecher, R. Kleijn, and G. J. Kramer, “Recycling potential of neodymium: the case of computer hard disk drives,” Environmental science & technology, vol. 48, no. 16, pp. 9506-9513, 2014.
O. Abdel-Hamid, A. R. Mohamed, H. Jiang, et al., “Convolutional neural networks for speech recognition,” IEEE/ACM Transactions on audio, speech, and language processing, vol. 22, no. 10, pp. 1533-1545, 2014.
R. Tsai, C. Đ. Nguyễn, Đ. S. Hồ, et al., “Using mobile multimedia platforms in teaching dental diagnosis,” Journal of Taibah University Medical Sciences, vol. 15, no. 4, pp. 265-271, 2020.
Y. Ying, “Application of flipped classroom teaching mode based on MOOC in modern educational technology teaching,” Journal of Computational and Theoretical Nanoscience, vol. 14, no. 2, pp. 1075-1078, 2017.
L. C. Ehri, and B. Flugman, “Mentoring teachers in systematic phonics instruction: Effectiveness of an intensive year-long program for kindergarten through 3rd grade teachers and their students,” Reading and Writing, vol. 31, no. 2, pp. 425-456, 2018.

Application of Android Voice Assistant based on Parallel Storage in Multimedia English Teaching Platform

Status:

Journal Publication

Version 1

Abstract

Figures

1. Introduction

2. Related Work

3. Technical Basis Of Parallel Storage And Android Voice Assistant

3.1. Parallel storage system structure

3.2. Technical basis of Android voice assistant

4. Design And Implementation Of Multimedia English Teaching Platform

4.1. Multimedia English teaching platform design

4.2. Realization of multimedia English teaching platform

5. Conclusion

Declarations

References

Status:

Journal Publication

Version 1