Yes.

Please start by introducing

Yes, you can begin now. Hello, is my voice audible?

Yes, your voice is coming through clearly. Please go ahead and introduce your professional journey in academia.

OK. Thank you. I'm Doctor Shiva. Currently. I'm working as an option professor in AC department at Amruta University. Bangalore campus Sir. I did my PhD from Saltar Vallabhai National. Technology Surat. I didn't take. Dominated Aravinda. During PhD my. Guide is doctor Suman Dev and during M tech time my guide is doctor Rajat Kumar. So I did my btech from Sanjay membrane.

OK. Thank you. I'm Doctor Shiva. Currently. I'm working as an option professor in AC department at Amruta University. Bangalore campus Sir. I did my PhD from Saltar Vallabhai National.

Okay, so you're currently an assistant professor in the ECE department at Amrita University, Bangalore, and you did your PhD at Sardar Vallabhbhai National—sounds like there's more to your story. Go on.

So, Dr. Shiva, thinking about your teaching at Amrita, what's one foundational concept in your field that you especially enjoy introducing to new students?

OK.

OK. During a. Under the university I used to teach a Siemens system or signal processing to the students. And my research field is a signal processing and especially in the signal processing. Speech recognition or speech emotion recognition? Yes.

OK. During a. Under the university I used to teach a Siemens system or signal processing to the students.

Alright, Dr. Shiva, you mentioned your experience in the ECE department. When you introduce a foundational concept to your new students, what strategies do you use to make sure they really grasp it, especially if it’s something they might find abstract or challenging?

Alright, so you taught signal processing. When you first introduce that topic, what do you do to help students who might be struggling with the abstract math or concepts?

Right, so you focus on speech emotion recognition in signal processing. When you first introduce that concept, what's one example or analogy you use to help students relate it to something they already understand?

Yes. In 2014, when I joined GITAM as a national professor there, I taught this signal system subject to student. Later these are in Amrita when I joined 2000. UH-25, September. I taught this subject.

Okay, so you started teaching signal systems at GITAM back in 2014 and later continued at Amrita in 2025. When you introduce signal processing to a new batch, what’s the very first real-world example you use to make the topic relatable?

1st signal. It's very important because everything is a signal. So I introduced first what is signal. And how signal is very relevant and important in our life. Because. Sigma is nothing but a physical quantity that varies with some. Independent variable and contains some information. And whatever speaking and doing is. All about inform. Is very important. The first I introduced about signal. Then I. Started system. Then I introduce about different type of system. Then I continue my subject.

Dr. Shiva, I noticed your recent publication, “Speech emotion recognition using MFCC-based entropy feature” in Signal, Image and Video Processing. How did you take the ideas from that research and adapt them for teaching your students, especially those who might not have a strong background in signal processing?

OK, uh. It's a very simple, it's a single procedure is a speech signal processing. Actually, uh. Emotion is nothing but emotion is defined as a strong feeling that can be created because of many reasons like most thinking, process, circumstance, behavior. Personal and professional relationship and many more. And it's very important in our life to detect and. Find the emotional condition of human being to understand the. That's why what I did that in my recent publication I found that MFC is one of the. A high use spectral feature for in the field of speech processing because it captures the. Like all the. Aspects or attributes of the speech. So I use MFCC. Signal metrics are. Of is speech signal then use. Column of the MMC symmetric to find the features. The main reason behind that is that the. Emotes is varying with respect to different frequency bands of the speech signal. So while we are finding the MFCC future, the MFC feature with respect to different segments or frames are also changing. So we can now capture the. The different attributes which captures the inform. Question about emotion by. Taking the columns of the MCC matrix, so I use the MFC matrix. And. I use the name of symmetry to find the features. Which is more relevant and emotional and primitive and I use that feature. To classify the speech emotion using. Deep learning class way like. DNN and I found that a good result. Though now those are not having signal processing proper background. Also you can say that emotion is varying with respect to different frequency band and to capture different frequency band based features I use the MFC Future. Which captures the attributes of the speech signal then. Those attributes are useful for recurrence, emotion, condition of human being. This is a simple concept I used. And. Yes.

Okay, so you broke down the MFCC approach for emotion recognition and tried to relate it back to how emotions show up in different frequency bands. When you explained this to students with little signal background, did you ever run into confusion, and if so, how did you address it in class?

OK, it's not pure single processing breakdown background required. Because, uh, yes, it's uh, we can use this, uh, by using Python. Language. I've taken a speech signal and 1st I found the MFCC matrix by using Python. Then I use that each column of the MFC metric to extract the feature like approximate entropy. Spectral entropy, permutation entropy, and mean. Then I use different combination of feature to classify speech immersion using. A deep learning algorithms like DNA classifiers. That's very interesting. So while teaching this all this thing I said overall. It's work And why is that required? Then I. Use this up now. Model to train and test it and if I'm good results students are feel very happy.

So you emphasized Python for extracting MFCC features and used entropy measures to help students connect the math to real model outcomes. When you demonstrated this in class, did you have students run the code themselves, or did you primarily show the results and walk through the process?

Yes Sir. Well, actually before that we started the Python concepts to store and. This would understand the Python language. So we've taken so much like before 1 semester itself. They learn about Python And Matlab Sir. They should know that the basic coding tactics so that. They can do the coding by themselves. Then after that, once they know the coding techniques, then I explain how to do. First I did the code, I explained the logic. And how we can write the program for this particular MFCC based feature? Then it's student taking some time. They try to write by themselves but their mistake, but I was there to guide them and letter what happened. They got the good result. And the. So much confident. Now they are very sure that they can do. This type of program and they can. Using specific. Same thing we can do with metalab also, but I use Python. To do this.

Makes sense. Let’s talk about your research next. Given your ongoing work in speech emotion recognition, where do you see the most promising, fundable research directions over the next few years—and which grant sources or funding agencies would you target to support that work?

Yes, speech. Emotion. Changing till now. On the 1 * * 6. And I had taken only single source like space. But so many others resources are available like easy signal is there. Um, video ceiling is there, image signal is there, and combinations from more than one that is known as. Multimoral signal sign. Now first of all I want to explore. The combination of more than one signal, that is multimodal signal. Are to extract future and use it to classify the speech emotions is the first target. Once it achieved that, then I can, uh, think that we can go for the real time of, uh, emotion recognition because we know that. A space emotion having many applications in the real world. Like improving the human machine interaction, it is useful. In the health and medical monitoring system and it is useful in the like call center, service center and to avoid accidents to find. Online education, marketing and many more. For great contents like parent entertainment and gaming. So because of many advantages. Though I want to explode in real time so that people get a complete benefit of that and even they can use their mobile to know the. Motion condition and. They can get out. I wanted to end up you. Used in the many applications as I explained earlier.

Yes, speech. Emotion.

You started to say "speech emotion" — can you elaborate on which aspects of speech emotion recognition you see as most likely to attract funding, and which specific agencies or grants you’d go after to support that research?

So you're hoping to expand into multimodal emotion recognition, combining speech with other signals like video or physiological data, and then move toward real-time applications—especially for things like healthcare and education. Can you give me a sense of which specific funding agencies or government programs you’d target for this multimodal research, and why you think your work aligns with their priorities?

Okay, till now I nothing in my mind that first I want to develop the moral then I will see. Umm, it's a real name project of government is there then I can try for that. If I'll get, if I'll get, then I. Go for it for real time work. Until now means uh, I not. I not try any government Funding Agency anything.

Given your background, if you couldn’t use slides or traditional lectures, how would you design a hands-on image processing lab session that keeps a large class genuinely engaged and learning?

People are interested when they will see something and emails posting nothing, but they can visualize it. So when like I can take the example of spectrogram, when you take the spectrogram they can visualize. The time frequency representation of the signal. Then they will find more interest. Obviously when people will see, they will find the interest. And when we see like spectrogram then dead. Yourself they can find that how emotion actually vary with respect to. Uh, frequency and. Then they can like that so many time frequency analysis there we can take in the form of emails. And that image is interesting and. Again, process that image to find the emotion recognition. Like. Similarly, this is with respect to my research if you take. Simple in image processing also same thing. When you see the images and they can process it, they'll find the. More and more interested. So when they'll find the interest means they can try to do something. Uh more instead of uh means one division signal processing where so many mathematics is there. And they cannot visualize it. But in e-mail. OK, we can say that they can visualize it. They can. Whatever the everything, they will see. And uh, they can try to achieve.

Alright, so you focus on making image processing visual and hands-on—using spectrograms and letting students actually see the results, which helps them get interested and try things themselves. If some students are still disengaged or just go through the motions, how do you track whether your approach is really working for everyone?

Yes, Sir, it's happened. Obviously means uh, like, uh, maybe some students are not good in programming or maybe, uh, they're not finding the exact information on how to capture. Through the image at that time I have to give special time or some extra time to them. I can discuss. When I discuss with them, they can say that really they can share their problem, whether they're not understanding the program or the. On that, understanding the concept of emotion recognition using uh. Time frequency analysis like spectrogram or triple transforms. Once I didn't find the problem, then I can spend some more time with them and I can try to make them understand. Once they will understand sure that. Uh, they can do it. It's not a big problem.

Alright, Professor, let's shift a bit. Accreditation cycles can be demanding — say you joined our department and found that outcome assessment data was inconsistent across courses. How would you approach fixing that?

OK, can you repeat this? On this question again.

Sure, I’ll repeat the question. Accreditation cycles can be demanding — if you joined our department and found that outcome assessment data was inconsistent across courses, how would you approach fixing that?

1st I have to find the problem where it is actually. Once he'll find this then I can work on that to fix it.

Alright, so you'd start by pinpointing exactly where the inconsistency is before working on a fix. Can you give me a concrete example of how you'd go about finding the root cause—would you look at rubrics, talk to faculty, or start somewhere else?

Yes. Obviously, we know that all people are not expected on all topics. If I'm finding anything difficulty, I have to find the professor or the faculty who is. According that I can spend time with them, I can take the help. And I've tried to fix it.

Yes. Obviously, we know that all people are not expected on all topics. If I'm finding anything difficulty, I have to find the professor or the faculty who is. According that I can spend time with them, I can take the help. And I've tried to fix it. Meanwhile, I can take the help of. For now, a number of resources are present. Like in Internet, I can try to try to fix it.

Let’s consider a situation you might face: Imagine a student files a formal complaint, alleging your grading was biased, but at the same time your department head is pressuring you to improve pass rates. How would you handle this conflict while maintaining academic standards?

Okay, so you’d connect with colleagues who have experience in the area and look for online resources if needed. Let’s consider a situation you might face: Imagine a student files a formal complaint, alleging your grading was biased, but at the same time your department head is pressuring you to improve pass rates. How would you navigate that?

OK, see it's it's happened. Generally when you deal with students there means quite young people. So we have to spend time. Uh, I'm thinking that the only one thing is that we can spend. Time and we can discuss with student all this thing. Uh, then, uh, slowly, what happened? I mean, sometimes. They think that the prophecy is. Yes, but. You have to say. You have to say the reasons of everything. Where they lose the mark, why they lose the mark and how means others are getting more means. What is the reason behind that? We can motivate it. Because in this time of the negativities is common. Everyone though, we have to try to convert that negativity to positively. And 1st we can. Interact more and more with the student. Then I think once they will understand that they can work hard and they can work on that slowly they will improve. And once they will improve means obviously. The pressing will be. Increase not required to improve something like it's like Bioshock and something not required to give extra marker because hod thing. You have to work and you have to make the student work so that. Thing will be automatic, the past and the personage will increase.

OK, see it's it's happened. Generally when you deal with students there means quite young people. So we have to spend time. Uh, I'm thinking that the only one thing is that we can spend. Time and we can discuss with student all this thing. Uh, then, uh, slowly, what happened? I mean, sometimes. They think that the prophecy is. Yes, but. You have to say. You have to say the reasons of everything. Where they lose the mark, why they lose the mark and how means others are getting more means. What is the reason behind that? We can motivate it. Because in this time of the negativities is common. Everyone though, we have to try to convert that negativity to positively. And 1st we can.

You mentioned explaining to students exactly where they lost marks and working to turn negativity around by spending time discussing their concerns. But what if your department head disagrees with your approach and insists you simply raise the grades to improve pass rates—how would you respond?

Alright, let’s look at how you’d connect your academic work to broader industry outcomes. Given your background in image processing and communication systems, do you have any industry contacts or collaborations that could help translate classroom learning into internships or placement opportunities for students?

OK. Actually I don't have contact of any industry till now but when I did my PhD my guide is there Doctor Suman they when so many I. I came across among many professors. So is that in the 1980s? Surat and they're working on. With collaboration of industries, uh. OK. So when I joined means I will just contact with them and try to get how to. Contract with the industry people who once will get. Interact with the industry people. That I can find the problem and I can. Save this with my store and I'm trying to solve it. So that we can approach. Proper route and proper way. Through college and we can start collaboration work with the requirement of the industry.

Given your experience with both Python and Matlab for signal and image processing, imagine you’re guiding a student through a project that involves embedded systems and real-time communication—what’s one specific challenge you’d expect them to hit, and how would you help them work through it?

Actually my work is on signal processing, especially in the speech emotion recognition. I never miss work on the embryo system. But it's not typical, they can learn it and uh, according to the requirement I can develop the Python programs and models. And help. Uh, specifically when you will do the projects, not it's a whole work to do at a time. We have to divide the project into small number of models and we have to work on the each model when we succeed for each model and then we combine all to get the success of the whole project. Now with with respect to real time communication and IoT projects or embedded system project. Based on the project we have to divide. It and we will work. Both the student and myself will work and explore it. Solve the problem by using. Either Python or MetLife according to the availability of resources.

Yes, is is mainly the fear. Uh, main problem is uh. Uh student before starting anything uh, they're having since uh, having fear our students are not good enough for. So we have to spend time, we have to explain that OK, first this is the small model we can divide like 1 project is that you can divide into four part. Then first the. We main target is that like one month time or two-month time is there. First we go through the basics then first we complete the small model. Once the first will come in, they will be. Get confident, then go for the second and third, then fourth like that, then we can combine. In this way means slowly we can build a confidence among student but once. I will succeed in the even in the first quarter then not required to boost up more they will able to do. The rest part by themselves. Meanwhile I have to. Work with them because since the areas are not mined. And main area is just completely different. So I have to work. I will to learn this to render so that it is useful for both and you can. When you succeed means you can succeed both.

So you break down projects into smaller modules and work through each part, even if embedded systems aren't your main area. When you switch from Matlab to Python for signal or image processing tasks, what do you find are the biggest pitfalls students encounter, and how do you help them get past those?

Alright, you emphasized building confidence by breaking projects into smaller parts, especially when students are new to a topic. Shifting gears, when it comes to evaluating students—like setting exam questions or grading lab work—what do you do to make sure your assessments are fair and actually reflect what students have learned?

Yes, that question should be practical oriented and that should not be solved in the live systems. When the student can think. OK, most surely you can allow them to do for openness exam so that they can think they can use the. Even a resources of the like journalist records what they did in the exam. Once they can find out. And then we can see that once you do the open SSM exam means student has to feel free and they will think that OK. Umm, the frequent is, umm, giving all opportunity, all the things to do work. Uh, still I'm not uh, able to do means what I will, uh, mark, so he will give. I deserve it up. Or if if I will do? And I can succeed me so. Obviously I think he will give more marks, so in this way the fairness will come. This is all we have to discuss the student. We have to spend time with the student. So because the maximum problems are happening with the. Communication when you're teaching me so. We're not a professor there, so we can. Deal with the young minds. When you're dealing with young minds means you have to spend time with them. Uh, what is going on in their mind? Once you will understand that means I think your, uh, things will be sort of. Uh, it will take some time. But in yourself. Throat.

Yes, that question should be practical oriented and that should not be solved in the live systems. When the student can think. OK, most surely you can allow them to do for openness exam so that they can think they can use the. Even a resources of the like journalist records what they did in the exam. Once they can find out. And then we can see that once you do the open SSM exam means student has to feel free and they will think that OK. Umm, the frequent is, umm, giving all opportunity, all the things to do work. Uh, still I'm not uh, able to do means what I will, uh, mark, so he will give. I deserve it up. Or if if I will do? And I can succeed me so. Obviously I think he will give more marks, so in this way the fairness will come.

You mentioned open assessments and letting students access their own records during exams so they feel the process is fair. When you design these practical or open assessments, how do you make sure the questions are structured clearly so that students aren’t confused about what’s being asked?

You talked about open assessments, giving students real-world problems, and making sure communication stays clear so they know what’s expected. When you’re supervising a student research project—say, in image processing—what’s your process for helping them select a topic that’s both challenging and realistic for their level?

Is ADA. It's very difficult to find what you.

Is ADA. It's very difficult to find what you. And there is this area. **** his legging.

Yeah, picking the right research topic can be tough for students, especially in a field like image processing where the scope is huge. When you have a student who's really struggling to narrow down their options, do you have a method or framework you use to help them focus—maybe looking at available datasets, current trends, or their own technical strengths?

Alright, Dr., it sounds like you recognize how challenging it can be for students to pick a research topic in image processing, and you help them work through that. Now, thinking about publishing research—can you walk me through a concrete example of one of your own research papers that was published in a reputed journal? What was the core contribution?

Put my topic like speech, emotion decomposition I can say. Uh, like one topic is that, uh, it's uh. Uh, various little more decomposition based. Feature to recognize the emotional condition of human being. That as I said, that. Umm, the simple in the research area, uh, when I started, I found that. Many people use the different features to to. Find it. Now the emotional condition of human being. But no one uses the decomposing technique. At that time. So why decomposition technique is essential? We can say that. Uh, the emotion is varying with respect to different frequency bands. Though when you process the whole signal then we are missing the so many. Umm, attributes to them which contain the information about emotions. So what you will do that first we will decompose the signal into lofts. Some of the. With respect to frequencies the what will happen? We will get a different frequency band signals now. Now different frequency band that captures the different information about emotions. Now instead of using the whole signal if you. Use the different segments. Are different frequency bands of the signal to extract the future Then again say that the. Features will be more reach of information about the emotion. Uh, then that of the whole speech signal based features, that is the core idea. And I thought and I work on that and I decompose the specific number of. Sub signal said. Then I use the term. To extract the feature. And. I use different features like MFC based feature. And I use like entropy features and I use combination of both. To pacify emotion using deep neural network classifier and I found a good result on that. So in this way you will first understand the core concept. Before going for any resources. Then you can apply the logic of what to do. In this BMD based decomposition method I use that technique of decomposition of signal. Now then extension of future using decomposition signal instead of the whole and that. For for my work like that for any topic you have to. Just find the main base or we can find the concept very slight. Then we can go for it either for a feature based or moral based or anything that is a secondary. But you have to find the basics about that topic, then you can work on that. Then it is more useful instead for starting directly.

Put my topic like speech, emotion decomposition I can say. Uh, like one topic is that, uh, it's uh. Uh, various little more decomposition based. Feature to recognize the emotional condition of human being. That as I said, that. Umm, the simple in the research area, uh, when I started, I found that. Many people use the different features to to. Find it. Now the emotional condition of human being. But no one uses the decomposing technique. At that time. So why decomposition technique is essential? We can say that. Uh, the emotion is varying with respect to different frequency bands. Though when you process the whole signal then we are missing the so many. Umm, attributes to them which contain the information about emotions. So what you will do that first we will decompose the signal into lofts. Some of the. With respect to frequencies the what will happen? We will get a different frequency band signals now. Now different frequency band that captures the different information about emotions. Now instead of using the whole signal if you. Use the different segments. Are different frequency bands of the signal to extract the future Then again say that the. Features will be more reach of information about the emotion. Uh, then that of the whole speech signal based features, that is the core idea. And I thought and I work on that and I decompose the specific number of. Sub signal said. Then I use the term. To extract the feature. And. I use different features like MFC based feature. And I use like entropy features and I use combination of both. To pacify emotion using deep neural network classifier and I found a good result on that. So in this way you will first understand the core concept.

So you took a decomposition approach in your speech emotion recognition research, focusing on frequency bands to extract richer features for classification, and you tested combinations like MFCC and entropy features with deep neural networks for better results. Switching gears a bit—when you’re explaining complex research like this, how do you make sure your students are really following along and not just nodding along?

Alright, you talked about your speech emotion decomposition research and how you approached feature extraction. I'd like to ask about your hands-on experience with image processing—can you give me a concrete example of an image processing project you've worked on, mentioning the techniques or tools you used?

OK, actually. Peace, emotion, recognition. I not work on image processing. But OK during my undergraduate term means during Mtech time we did some work on image processing but. Oh, after that I never went for any immense processing based Pepper publication. Yes, but we can say that the image processing means. Inventive work already did that. Instead of taking a one time sense signal I take. Time frequency based images. And a process that images to classify emotion using. DNN classical like CNN conversion neural network. That's a but I not. More about image processing, but it's the area now I want to work. Because the speech I did, I want to work with image and I want to combine both speech and image. To get a better classification performance.

OK, actually. Peace, emotion, recognition. I not work on image processing. But OK during my undergraduate term means during Mtech time we did some work on image processing but. Oh, after that I never went for any immense processing based Pepper publication. Yes, but we can say that the image processing means. Inventive work already did that. Instead of taking a one time sense signal I take. Time frequency based images. And a process that images to classify emotion using. DNN classical like CNN conversion neural network.

Right, so for your image processing experience, you mostly worked with time-frequency representations as images for emotion classification, using CNNs rather than traditional spatial image processing tasks. Let's switch gears—when you communicate complex technical topics to students, how do you make sure your explanations are structured and clear, especially for students who might struggle with the basics?

You mentioned some exposure to embedded systems but not a lot of hands-on work. Can you tell me about a time you actually had to implement or troubleshoot a communication protocol or work with embedded hardware, even if it was a small part of a larger project?

You go for it. Obviously nowadays it's very interesting to explore new areas and it's challenging also. So these emails posting embedded and IoT all this thing in future. First I will finish my what is the penning works on the? Space processing. After that I will explore all these parts also. And I will include in my recent work. So that me and my students will get benefit for that. Benefit by using this.

Okay, so you're planning to build more experience with embedded and IoT as you finish your current speech processing work. Shifting gears, when you're presenting a technical concept to students—especially something complex—how do you make sure your explanation is clear and structured so that everyone follows along?

Yes Sir, before going to the class it is very important to do the work on that. OK, because whatever the complexity, first you to. Uh, decompose and try to make it simpler so that the student can understand. Uh, for that you have to. Now work more and more and you have to think like a student, not like a professor. And where the student really stuck, we have to work on more on that. Once you'll understand that, then it is easy to. Explain them, still some student not able to follow. You have to find some more time or it required you to repeat that topic one more time. Before 1 is necessary otherwise if you only have been facing problem. Then we can say that. Uh, uh, obviously. Uh, we can spend time only that particular student and try to explain them so that they can get. My understand and get. Benefit of that. Yes.

Right, you emphasized thinking like a student and breaking topics down before class so your explanations are clear. When you’re preparing a lecture or lab, how do you structure your delivery to keep it organized and make sure students don’t get lost along the way?

My specialization is a speech based applications. OK, so when we know that maximum prob things are related with speech it may be diseases or maybe emotion. Many things you can extract from this. Speech because the speech is a like a natural phenomenon given by the God.

Alright, we're wrapping up. Is there anything you'd like to ask about the role or the team before we finish?