Hello all,

This is a very new and fast growing research area. A high end DSP application with a lot of computer science and engineering and mostly web application. Though content based multimedaia can come to hand held and mobile devices, may not be in a near future. But what can happen is that, these are available as a web service and a mobile device may consume this service.

So, what is multimedia data minning? Unlike the usual text based search, here the data is gathered from the very content of an audio clip or an image or say a video.

e.g say an image contains a ball. Then even with the existing image processing algorithms the ball can be detected and stored with a tag ball. So that people can search for an image that has ball in it rather than searching for a image that has a file name 'ball'.

Similiarly, one can search for an audio clip with a word, typed or spoken. Or one may search songs in the compuetr or on the internet by singing or even humming the song.

Basically, to do all this, one must be able to know what to look for in the image or audio and then have an accurate representation that can be relied on.

## Wednesday, May 7, 2008

## Friday, May 2, 2008

### An Introduction to Linear Algebra

This part is not so mine. But I do wish to put it here, so that It may be beneficial to somebody. This is only an introduction in layman language. This should encourage you to go on and take some books on this subject.

I wanna tell you here is not what my friend had to say in the DSP forum when he took a session. He did a good job only that what he talked was irrelevant to the audience. My idea as initiator of DSP forum in college and in here as well is to make DSP an perception, an intuitive feel rather than just mathematics. Mathematics is to substantiate what we talk and feel. It is the final step and not the first step in solving the problem, at least I wanna think so. That makes my life easier.

So, back to linear algebra.

Linear algebra, more specifically, vector spaces, I would like to think of it as only generalization of our usual understanding of algebra. That's all. Nothing more than that.

So, where do we start.

Lets talk about an N-dimensional euclidean space. Am i already going too fast. Well, By, "N-dimensional euclidean space" all that I mean is a Cartesian Co-ordinate system like system of x, y, ... co-ordinates, N such axises. Each axes is orthogonal to all others. "Orthogonal" just means normal or 90 degree. The generalisation is that the dimensions need not be like Cartesian. May be polar, a 2-dimensional euclidean space or extension of it as a cylindrical or spherical co-ordinate system.

I will not go into any formal definition of vector spaces, or for any other term that I would use. This is not the place for the formalities.

Then, having a space, we can visualise a point in this space, like in a 2-d plane. We can as well imagine lines or planes and volumes. If there are morfe than three dimensions we, call it a hyper object as in hyper-line.

This in essence an account of vector spaces in layman's language. We can extend most of the theorems in the regular school algebra with possibly a litle modification or some assumptins.

I wanna tell you here is not what my friend had to say in the DSP forum when he took a session. He did a good job only that what he talked was irrelevant to the audience. My idea as initiator of DSP forum in college and in here as well is to make DSP an perception, an intuitive feel rather than just mathematics. Mathematics is to substantiate what we talk and feel. It is the final step and not the first step in solving the problem, at least I wanna think so. That makes my life easier.

So, back to linear algebra.

Linear algebra, more specifically, vector spaces, I would like to think of it as only generalization of our usual understanding of algebra. That's all. Nothing more than that.

So, where do we start.

Lets talk about an N-dimensional euclidean space. Am i already going too fast. Well, By, "N-dimensional euclidean space" all that I mean is a Cartesian Co-ordinate system like system of x, y, ... co-ordinates, N such axises. Each axes is orthogonal to all others. "Orthogonal" just means normal or 90 degree. The generalisation is that the dimensions need not be like Cartesian. May be polar, a 2-dimensional euclidean space or extension of it as a cylindrical or spherical co-ordinate system.

I will not go into any formal definition of vector spaces, or for any other term that I would use. This is not the place for the formalities.

Then, having a space, we can visualise a point in this space, like in a 2-d plane. We can as well imagine lines or planes and volumes. If there are morfe than three dimensions we, call it a hyper object as in hyper-line.

This in essence an account of vector spaces in layman's language. We can extend most of the theorems in the regular school algebra with possibly a litle modification or some assumptins.

### The Second Time

So, at the end of first session we realized that We cannot continue this the way initially thought about. But to sease our problems someone from the 2nd year batch said he will take a bridge course for the freshers in signals and systems.

That being the background, the second session started with rather a buzz. My immmediate juniors dint want to attend this session coz,they thought it was a preliminary detail. But I said his session was only an half an hour one and i would continue later. I basically wanted some audience.

But The things were not to go as we planned. This guy talked everything but signals and systems. We expected him to talk about analog and digital signals, representation, systems perspective, stability issues, Linear Time invariant systems, the difference and deferential equation representation and so on. This guy talked about Linear algebra of all the things in the world. These poor freshers, who are new to even matrix algebra, to my surprise were listening carefully. I will elaborate on what he talked in another blog, but for now it is not so important.

Finally, he concluded and expressed a wish to talk about uses of differentiation. I told him he did a good job but we shall stop there. Already an hour was over, I asked people if they are ready to bear me for half an more, and I promised them I wont bore them.

And there starts a wonderful journey of DSP forum... Continued...

That being the background, the second session started with rather a buzz. My immmediate juniors dint want to attend this session coz,they thought it was a preliminary detail. But I said his session was only an half an hour one and i would continue later. I basically wanted some audience.

But The things were not to go as we planned. This guy talked everything but signals and systems. We expected him to talk about analog and digital signals, representation, systems perspective, stability issues, Linear Time invariant systems, the difference and deferential equation representation and so on. This guy talked about Linear algebra of all the things in the world. These poor freshers, who are new to even matrix algebra, to my surprise were listening carefully. I will elaborate on what he talked in another blog, but for now it is not so important.

Finally, he concluded and expressed a wish to talk about uses of differentiation. I told him he did a good job but we shall stop there. Already an hour was over, I asked people if they are ready to bear me for half an more, and I promised them I wont bore them.

And there starts a wonderful journey of DSP forum... Continued...

## Wednesday, April 30, 2008

### DSP Forum Started

When GP ma'am proposed that a new forum will be operated where seniors get to interact with juniors and teach them and learn with them.

It was difficult choosing DSP, but me and Sandy chose it, coz off late we were working on few algorithms classified as statistical signal processing. We thought Matlab will fill the gaps in our attempt, and we said its DSP with MatLab

The first day, Principal came and announced full support. And officially, DSP Forum was flagged off.

The night before the very first class, I took out Simon Haykin, "signals and Systems" book. With the intension of brushing through I made some quick notes for the next day. To our surprise there were more freshers than our immediate juniors.

They said right right when they said, we got to come out of our comfort zones to get better. I was caught off guard when I saw a crowd of first years, 6 months into engineering looking at me with a hope that I am gonna teach them DSP and MatLab.

That day I did a descent job. I chucked my notes and started talking impromptu. I talked about Sampling, I stated Nyquist theorem and I talked about Digital systems. In fact I was able to put in place the notion of signal and systems. Borrowing the idea of Infinite loops and non converging infinite series I could drive the idea of instability in systems. I ended the class with few feed backs from the students but I knew I was short on what I wanted of this class.

It was difficult choosing DSP, but me and Sandy chose it, coz off late we were working on few algorithms classified as statistical signal processing. We thought Matlab will fill the gaps in our attempt, and we said its DSP with MatLab

The first day, Principal came and announced full support. And officially, DSP Forum was flagged off.

The night before the very first class, I took out Simon Haykin, "signals and Systems" book. With the intension of brushing through I made some quick notes for the next day. To our surprise there were more freshers than our immediate juniors.

They said right right when they said, we got to come out of our comfort zones to get better. I was caught off guard when I saw a crowd of first years, 6 months into engineering looking at me with a hope that I am gonna teach them DSP and MatLab.

That day I did a descent job. I chucked my notes and started talking impromptu. I talked about Sampling, I stated Nyquist theorem and I talked about Digital systems. In fact I was able to put in place the notion of signal and systems. Borrowing the idea of Infinite loops and non converging infinite series I could drive the idea of instability in systems. I ended the class with few feed backs from the students but I knew I was short on what I wanted of this class.

Subscribe to:
Posts (Atom)