• 1 year ago
चंडीगढ़ की पंजाब यूनिवर्सिटी में एक ऐसा कमाल का साॅफ्टवेयर बनाया गया है जो एआई आवाज़ का पता लगा लेता है.

Category

🗞
News
Transcript
00:00AI where life changes, there are days and days of such tools, which are easier for life.
00:09In the case of Punjab University, Anthropology Department has a new advice that has been made for the right tools.
00:21And this team has proven that AI's voice can be changed by the actual voice and how it can be generated by AI.
00:36Professor Keval Kishan, let's talk about it.
00:39Sir, in the situation, you have confirmed that AI is not only a tool, but also an easy life.
00:48When you look at the time, the challenges are also being cleaned up.
00:55Sir, tell us that you have created a new voice, a new tool, a software,
01:03and it tells the AI to generate a difference.
01:10So tell us when this idea came and how did it work?
01:15Madam, this tool is called Voice Identification Tool.
01:20In the Anthropology Department and Forensic Science, we do all the human body with the whole body.
01:26But the voice is also a part of human body.
01:29So we have thought that first of all, we have done a new work.
01:34Human body, human morphology, face, hair, human body, gait, child, we have done a work.
01:40We have done a work that has a societal value.
01:44When we thought that AI-generated voices, many people are concerned,
01:52and many cases have come, and we know that AI-generated voice is possible.
01:56Why can't we understand AI-generated voice?
02:01And we have done a data collection.
02:04Now, we have been collecting the previous data from the Master's Student's Master's.
02:06I made this project to test the students' mantle,
02:08and originally data collected,
02:10we have collected hundreds by hundreds of people,
02:13and used to have been completed by the original voice.
02:14And then I have cloned hijack through AI,
02:17which is called AI-generated voice,
02:21We have created the software in the software that we have created the AI-generated voice and the original voice.
02:38We have reached our program to 80% of our accuracy.
02:44We have identified it as well as the government of India has given it on copyright.
02:51We want to elaborate more on this data.
02:54We can collect more data from 90% to 92%.
02:59We can use this tool in the market also and we can give this tool to the investigating agencies also.
03:09We can investigate it and the police and law enforcement agencies will be easy to do.
03:16Our forensic scientists' work is easy to do.
03:19We need to do easy work and criminal work is easy to do.
03:24We can do easy work and use the AI-generated voice that is used to be easy to do.
03:31We have to use the AI-generated voice to remove the AI-generated voice.
03:36We have tried to use the AI-generated voice and use the AI-generated voice.
03:39It is easy to use the AI-generated voice voice-over.
03:43I would like to tell you that this software can be solved with this software, which can be resolved with this software?
04:00Yes, I would like to tell you that this software is like this, which distinguishes AI-generated voices and original voices.
04:08When we have a call or a fake or suspicious call, we will immediately report.
04:17If the case is like this, and the police know that this software is like a software,
04:26then it will not take time within 10 minutes.
04:29We can tell us whether this is an AI-generated voice or this is an original voice.
04:35This is an AI-generated voice.
04:42But AI also works in supervision.
04:44It is not that the forensic scientists and police personnel are sitting,
04:50they are also working on their own.
04:53This is a practical use of this model.
04:58This is why criminals should not be afraid of this work.
05:02What do we need to do now?
05:03Yes, sir.
05:04One more thing you can see is that when people have online calls,
05:09which are FaceTime or video calls,
05:12there is a situation that there is a person sitting there,
05:16or a judge sitting there,
05:18or a top officer sitting there,
05:20and he is inquiring with them.
05:21So in this situation,
05:24what do you have to do with your voice recorder,
05:27or voice software,
05:28can you help,
05:29or deduct,
05:30that it is an Indian officer,
05:33or a judge of this officer?
05:36Or,
05:37or a judge of this officer?
05:38Yes,
05:39it is not in software,
05:40but if the voice is a case,
05:43if he is talking about something,
05:45or if he is talking about original voice,
05:48then when you do a role,
05:50or something else,
05:51we do not have to work on it.
05:53It is possible that some days,
05:54if we do a child,
05:55or a PhD candidate,
05:57or a candidate,
05:58who is working on it,
05:59then it is possible to work on it,
06:00that we can do work on it,
06:03and that we can do it.
06:05It is a good idea,
06:06that we will work on it,
06:07but we have not done it,
06:09until we have to do it,
06:10that we can do it,
06:11or we can do it,
06:12but we can do it,
06:13or we can do it.
06:14So,
06:15the voice of this thing is,
06:17voice, we know that AI is generated, then we can give reports that AI is generated.
06:24Let me give you an example, like YouTube, there are a lot of voices.
06:29Today, there are such voices that are AI generated.
06:32People want to talk about the voice in Amitabh Bachchan's voice.
06:35However, Amitabh Bachchan's voice is not originally.
06:38We think that Amitabh Bachchan's voice is originally.
06:41But, we can see whether this belongs to Amitabh Bachchan or not, or AI generated voice.
06:49So, we can do such a thing, madam.
06:52There are a lot of videos on YouTube, like foreigners who are a big motivational speaker.
06:58They are using their voice.
07:02And they are motivated.
07:05But it is not the actual person's voice.
07:10They are just generated.
07:12And in this case, there are such things that the singer, the best singer,
07:18they are also using their voices.
07:21They are also using their voices.
07:23So, can we do such things?
07:28Yes, madam.
07:29It is possible to detect this voice.
07:31It is possible to detect this voice.
07:33It is possible to detect this voice.
07:34It seems that the voice of this voice is not possible
07:36to detect this voice.
07:38I know it must be one person who is not a members of the audience,
07:40but it must be one person who is not a member of the audience,
07:43but it would be the same person.
07:44It means that we have achieved something like that, because we have made a software, we have got a copyright, so it's a very good thing.
07:50It means that we have achieved something, but we have achieved something, and we can know what you want.
07:58I want to tell you that this is an anthropology department.
08:02There is a human mind of study, madam.
08:06So, we always want to do something new.
08:10This is a practical use of anthropology.
08:15We have shown that this is a practical use of anthropology.
08:22So, we have applied use of subjects.
08:26For example, in anthropology, forensic science, we have just seen what is the practical utility of the model.
08:33Practical utility is a good model, which is a societal work, which is necessary in society.
08:39The welfare of society is necessary in such a work.
08:43We have a scholar here who has done better work in this project.
08:49It is not the case of Kevil Kishin's perspective,
08:53but it is an experience of these scholars who are working on this software.
09:00Quail Kishin's perspective, despite this idea,
09:16what was the most important time about studying and how do weต do education?
09:22we can work on voice, so the master's girl, Pratibha,
09:27he gave the title that we will work on voice.
09:30So first of all, if voice, audio, video,
09:34then deep learning works very well.
09:37So we started doing some neural networks
09:40like CNN, ANN, but it was a short period
09:44because it didn't submit our dissertation
09:46and the data set was very low, like 50 individual data sets
09:50and we had to clone it, so deep learning
09:54is a little bit difficult.
09:56So we tried the machine learning models
09:58again, we extracted the features
10:00and created the Excel file
10:02and then, on the basis of the Excel file
10:04we tried to clone the original voice
10:06and classify it to 80% accuracy
10:10and we thought that it was a good accuracy
10:12because as much as it is,
10:14the accuracy was best.
10:16You talked about
10:18the data that you have collected
10:20from the data,
10:22which data has collected from the data
10:24from the data,
10:26which data has collected from the area
10:28and which people have participated in it?
10:30Yes,
10:32so this is the Pratibha area,
10:34Haryana,
10:35he collected from Haryana
10:36and he told us a text
10:38to read it.
10:40The age of 18 to 40,
10:42the age of people was the age of 18.
10:44And then, when the data has come
10:46to us, we had the same data
10:48which are freely available software
10:50to clone it.
10:52If anyone is a criminal-minded person,
10:54he can clone it easily
10:56and use them.
10:58So, for human ears,
11:00we felt that it was the same person.
11:02So,
11:04it has actually been an individual
11:05but it was an individual voice.
11:06So,
11:08when the data comes to investigation,
11:10the officer works on a level
11:12and the case
11:14will be very difficult.
11:16So,
11:18if this is a software
11:20used in which
11:22you can detect
11:24this?
11:25Or the AI voice is a real
11:26or an AI voice?
11:27This is a very recent software in the infancy stage, so now we want to get more and more data, so that this model will give it to 95% accuracy,
11:38because there is a problem solving crime, so we can't use any recent model in crime investigation, until and less it will be best fit.
11:47This is our case, so we will train more and more data set and represent it in the application form.
12:01These were the Panjab University Anthropology Department's scholars and professors,
12:06who have proved that AI is not just a tool, which can be solved with human challenges.
12:16We have seen some such issues, where the calls of people in their own lives were reported,
12:23they were manipulated, their families turned out on their own,
12:25their parents turned out to have money, their parents turned out to be made up of money.
12:30These are all cases that of police were found on social media,
12:37but most recently people inside the family are found on social media.
12:40to be very difficult to understand.
12:42The software that the Punjab University
12:44has created by the software
12:46is the way that
12:48is helping us to help us.
12:50In this case,
12:52there is not a direct investigation
12:54agencies to share
12:56but as the professor
12:58has told us,
13:00if they have more data
13:02then they will work on accuracy
13:04and if they can
13:06then they will share
13:08and share them
13:10so that they can use
13:12the software to use.
13:14Camera person
13:16Kishan,
13:18ETV,

Recommended