Journalof Name, Journal NetworksVol. and 1, Telecommunication Systems, Vol. 2 (1), 23-27, 2016 ISSN: 24698288 Published online: www.unitedscholars.net/archive
Speechmath IPTV’s Verification System Mahmood Alfathe, Veton Z. Këpuska Electrical and Computer Engineering, Florida Institute of Technology, Melbourne, USA
[email protected]
ABSTRACT In today's world, IPTV services are considered among the most prevalent technologies. One emerging challenge is the parental control. This paper demonstrates how to provide users with a mathematical quiz depending on the suitable program's age. The goal is to determine the answer from the user’s voice using the google speech recognition API. This work produces authentication results for two different age groups. All results are presented after viewer groups were tested with the same programs broadcasted over the IPTV service. Recommendations and a final decision are provided depending on the results obtained from this work.
Keywords: IPTV; authentication; speech recognition; ROC curve, FRR, FAR.
INTRODUCTION While cable tv’s, satellite tv’s, and terrestrial tv services were the dominant technologies that every tv user utilized to watch tv, a new technology was developing behind the scenes. This technology became a threat to the other tv services; and it is known as IPTV [1]. The growth of the internet, the variety of services provided within itself, ease of access and the ability of different devices to connect to the internet, made it of multi-purpose. One crucial implementation being the IPTV. However, a common problem was raised with such service, which is the parental control. As internet devices are widely available to a wide range of users, they became more exposed to children [2]. The development of the internet and in general personal devices made more technologies available to be used for accessing as well as
controlling the access to such devices, like microphones on phones or tablets. The most important advantage of such systems over password systems is its ability to exert more control over a system. Presently, for IPTV systems passwords are considered the weakest point. Most of IPTV operating systems use a four digit numerical password. Such a password can be easily discovered by guessing or using special patterns like birthday dates or special dates, or in the worst case using the trial and error, for up to 10000 tries [3]. The suggested system in this experiment uses different levels of mathematical quiz questions. The analysis of the recognized user’s voice should be the quiz’s answer. This experiment tests two factors: 1. The mathematical question. 2. Error rate. Can the suggested questions be a good test to verify the suitable age as a key to authorize the viewer? The error rate will be calculated in two phases: a. answers submitted by viewers, true / false answers, and b. speech recognition through the web application. How many tries until the web browser will translate the speech into the correct text?
METHODOLOGY In this experiment, the system will be tested and the produced False Reject Rate (FRR), and False Accept Rate (FAR) will be analyzed [6]. Improving the number of genuine users who have been rejected or accepted by error can improve the
© 2015 Mahmood Alfathe, open access article. Distributed under the terms of Creative Commons Attribution (CC BY) license 4.0.
Alfathe et al, Journal of Networks and Telecommunication Systems, ISSN: 24698288, Vol. 2 (1), 23-27, 2016
overall performance of such biometric authentication system. The Authentication process was developed to go through these steps: • Firstly the viewer will request a specific channel. The system will then analyze the Channel’s EPG to get the program's name broadcasted at that moment by Sending a request to (www.common sensemedia.com) [6] with that name. This is done to receive a detailed description of the program and the suitable age for viewing such a program. • Using the extracted age a decision will be made performing the following actions: If the program’s age is 5 & under, the channel will be broadcasted to the viewer (no further action will be needed). If the program’s age is between 6 & 11 then the system will request a question directly from “math.ly”[7] at a “Simple arithmetic” math level, as all questions consist of exclusively the four basic arithmetic operations. Finally, if the programmer’s age is 12 & older then the system will request a level-1 Algebra mathematical question from “math.ly”. In the described experiment the following was applied: A. Biometric used in this experiment The user’s voice was the Biometric used as an Authentication trait. It has been chosen due to the fact that it represents the most convenient Biometric trait that can be deployed with IPTV. Almost all IPTV devices, produced by commercial companies, started to use the user’s voice to perform such a search or to request a program. In this experiment, the voice will not be used to make a search nor to request a name for a program, instead it is used to answer a mathematical quiz question where the difficulty of the question is directly proportional to the age of the subject.
B. Data Base implemented in the project The first phase of this project was the task of picking the Database where the voice prints will be stored. This is a crucial step and makes sure that all voices used in this experiment are obtained under the same conditions, which is an important factor. In this project 3 groups of 20 users in each group were selected. The first group included tv viewers 4-5 years of age, the second group was for users between the ages of 6-11, the third group was for users that are 12 years and older. All answers were captured using the same microphone (Realtek High Definition Audio), all recorded in the same room. C. The Application used to implement the project The DVBlink software [4] was chosen to extract the EPG information for the terrestrial channels broadcasted in Brevard county Melbourne Florida. EPG extraction and parsing were programmed with Python. Finally, all the users' interactions were recorded with pages built using PHP & HTML languages. Furthermore, web pages including the mathematical quizzes retrieved from the internet were also used and aided in retrieving the suitable age for each IPTV channel. The IPTV service was operated under Apache web server.
SYSTEM BLOCK DIAGRAM The suggested system was built to perform a new channel authentication technology instead of the applicable numerical password used with the available IPTV systems. The system will go through the following phases as shown in Fig (1): 1. Channel request. 2. Channel program data analysis. 3. Expanding channel data. 4. Retrieving a suitable math quiz. 5. Requesting a vocal answer from viewer. 6. Compare the answer to the quiz’s correct answer.
24
Alfathe et al, Journal of Networks and Telecommunication Systems, ISSN: 24698288, Vol. 2 (1), 23-27, 2016
7. Making the final decision. •
•
• Fig. 1. System diagram
RESULTS AND DISCUSSION • •
•
The suggested system authentication procedure should impede minimally with the channel viewing and selection. The results were classified into two groups depending on the age of the user. The first group consisted of the viewers that fall in the 6-11 years old age group, while the second group was for viewers aged 12 years and older. Each group consisted of 20 viewers. Both groups were using the same system and requesting the same channels and programs. 12 viewers from the first group who represented 60% of the group, could pass the test and view the requested program.
While 8 viewers or 40% from the first group could not pass the test, as shown in Fig (2). The results for the second group showed that 80% passed the test, and 20% could not. Therefore, 16 users could correctly answer the mathematic quires, while 4 viewers could not. Fig (3) shows results for group 2. Fig (4) describes the first group's success cases and provides details about the failure cases. The accepted viewers are named True positive (TP). 5 viewers who represent 25% could correctly answer the quiz & the google API correctly recognized the speech. There were 30% of viewers rejected as they could not answer the quiz and the speech recognition recognized the speech as True Negative (TN). On the other hand, there were 5 users who were incorrectly accepted though they could not answer the questions correctly and the google API could not recognize correctly therefore they are False Positive (FP). The last 4 users who represent 20% of the total group were incorrectly rejected as they could answer the quiz but the google API failed to recognize the speech. They are labeled as False Negative (FN). Fig (5) illustrates the success and failure in detail for group 2. Group 2 shows different results as there was a very good enhancement with the (TP) results. There were 15 users, who represent 75% of the viewers, that could correctly answer the quiz and google API could correctly recognize the viewers’ speech. One viewer who constitutes 5% of the total could not answer the algebra question & google API correctly recognized the speech, therefore, the viewer was denied (TN). Furthermore, there was 1 viewer who incorrectly answered the quiz but the google API recognized a different answer and accepted the viewer (FP), he/she constituted 5% of the total group. Finally, there was 3 viewers, who represented 15% from the total group, that were incorrectly rejected as they could answer, but google API could not correctly recognize the correct answer.
25
Alfathe et al, Journal of Networks and Telecommunication Systems, ISSN: 24698288, Vol. 2 (1), 23-27, 2016
Fig. 5. First group, TP, TN, FP, and FN.
Fig. 2.Pass & Fail rate of first group
Fig. 3. Pass & Fail rate of second group
CONCLUSION Previously, the four digit numeric password was the only solution used for parental control for most of IPTV devices. The suggested application utilizes “age” or “realization” as a success factor to authenticate viewers. With the IPTV mathematical authentication the results showed a small loss with the genuine users. On the other hand, the number of imposters from the first group was improved as 60% of the viewers passed while the rest could not view the IPTV request. Google API speech recognition was a good addition to the system especially with the real time processing using the service provided by google and the future enhancements that google would add to its service. The only drawback with the google speech recognition API is the fact that not all the results were translated correctly especially with group 1 due to their age, as there was 20% of viewers who were incorrectly rejected and 25% of viewers who were incorrectly accepted, due to errors.
REFERENCES Fig. 4.First group, TP, TN, FP, and FN
1. Xiao, Yang, Xiaojiang Du, Jingyuan Zhang, Fei Hu, and SghaierGuizani. "Internet Protocol Television (IPTV): The Killer Application for the NextGeneration Internet." IEEE Commun. Mag. IEEE Communications Magazine 45.11 (2007): 126-34. Web. 26
Alfathe et al, Journal of Networks and Telecommunication Systems, ISSN: 24698288, Vol. 2 (1), 23-27, 2016
2. Fernández-Vilas, Ana, Rebeca P. DíazRedondo, and Sandra Servia-Rodríguez. "IPTV Parental Control: A Collaborative Model for the Social Web." Information Systems Frontiers InfSyst Front 17.5 (2014): 1161-176. Web. 3. Këpuska, Veton Z., and Mahmood Alfathe. "Improving STB Devices’ Parental Control." International Journal of Engineering Research and Applications (2015): 133-36. Web. 4. "Dvblogic.com - Dvblink." DVBLogic. N.p., n.d. Web. 16 Feb. 2016. 5. Jain, Anil K., Arun A. Ross, and KarthikNandakumar. Introduction to Biometrics. New York: Springer, 2011. Print. 6. "Common Sense Media." Reviews & Age Ratings. N.p., n.d. Web. 16 Feb. 2016. 7. Math.lyWe're SO BETA." Algebra Practice Problems: Linear Equations. N.p., n.d. Web. 16.
27
