Digital Image Processing

1. Image zooming with Interpolation.

2. Quantization.

3. Color Transformations: RGB component decomposition, YUV Color Transformation, YCbCr Color Transformation, YIQ Color Transformation.

4. Non Linear Filter: Design of the following types of filters: 1. Arithmetic mean 2. Geometric mean 3.Harmonic mean 4. Contra-harmonic mean 5. Median filter 6. Min 7. Max 8. Mid-point 9. Alpha trimmed mean filter.

5. Histogram equalization.

6. Design Inverse Gaussian Filter.

7. Design Wiener Filter.

8. Design Geometric Mean Filter

9. Blending Two images: Image Enhancement.

Neural Network

1. Implementation of Gradient or Steepest descent.

2. Implementation of Conjugate Gradient Technique.

3. One Step MLP using Steepest Descent, One Step MLP using Gradient Descent; Processing the results of MLP.

4. Classification error based on the Gradient Descent approach.

5. Convolution Neural Network using Keras using Google Colab.

6. Character Recognition using Convolution Neural Networks using Keras using Google Colab.

7. Coin and Scrap Recognition using Convolution Neural Networks using Keras using Google Colab.

8. Transfer Learning using Google Colab.

Digital Video Coding

1. Huffman Coding.

2. Golomb coding with prediction.

3. LZ77 coding.

4. Scalar Quantization.

5. Vector Quantization.

6. JPEG-LS (LOCO)

7. CALIC code.

8. Subband analysis.

9. JPEG Baseline.

10. JPEG Lossless.

11. EZW

12. SPHIT

13. 1:2 Sub Sampling and 2:1 Up Sampling

Malcomm Poster competition: Energy Sumit

Abstract:

Water is an essential commodity for everyday use and consumption. The importance of optimal and sustainable use of water, for day to day life, is crucial, and so is the importance of assessing the quality of chemicals used in each household and the after effects of that to the environment. The pollutants can also come from the farming, factory production and hence regulating the pollutants that are introduced in the water, hence, is also important in this sphere of life. In this novel production we concentrate to work on the closed circuit of household along with the water supply with future expansion to industry and farming kept in mind.

Deep learning algorithms have been deployed in measuring the quantity of water usage with each household unit. The water usage is also assessed for the pollutant, at the central Water analysis and assessment unit from which the data is send over to the Evaluation and prediction unit, where we deploy deep learning algorithms to predict the overall usage pattern and quality, by assigning a threshold based on family size, average income, previous water consumption record and time of maximum consumption. There will also be a set of data for the quality assessment data collected for threshold of the water quality assessment unit such as dissolved oxygen, pH, conductivity, and nitrate. Based on these data, we will create a reward system to provide incentive to the household to get better chemicals for day-to-day use and also to better manage the consumptions. There are water meters and sensors for the data procurement and processing, before and after the water leaves the household.

The water supply unit connect to the distribution unit, which will be fed with evaluation and prediction of each household and hence will have the data for the rewards, will appropriately distribute water accordingly. The feedback loop is kept to incentivize the household member to be conscious. This compensation could be designed as a bill pay reduction of extra water distribution. The Evaluation and Prediction unit uses Feed Forward neural network and along with the inputs

from the data storage unit and water analysis and assessment unit, comes up with a prediction score. The Water analysis and assessment unit has sub units for data manipulation and normalization, a classification and prediction unit for quality assessment and future prediction based on the old historic record, and finally an evaluation unit for water quality assessment score unit for assessing the reward. Once the final score has been evaluated and predicted, it is send over to the distribution centre. This closed loop feedback system provides individual household assessment and prevention criteria along with positive reinforcement in the form of award system.

Wireless Communication

Correct Identification of QPSK Signal in a deep learner, after the signal has been exposed to Rician channel and awgn noise added to it.

Versatile Home Automation System : Final Year Project Work (Bachelors of Technology)

Project Link: Versatile Home Automation System

Home automation systems are good example of the things which have integrated the applied electronics with our daily life in recent times. In our project, we have designed, implemented and tested such a system, which can carry out the task of home automation very efficiently, and the user interface is so versatile that the system can serve a wide range of purposes, both in home and industry. The system is an electronic circuit, which control different types of electrical loads like lamp, fan, TV, Computer, pump etc. The input command, which will be executed by the device, can be given in different ways: 1. Using Telephone: A cellular phone or GSM modem can be attached with this system, and the user can operate any of the connected loads by making a call to the attached phone, from any other phone from any part of the world. The call will be received by the device and the user will be prompted for entering the command through the keypad of the phone he is using. The command will be processed by the device, and the requested operation will be done. After the operation, the user will be informed through voice feedback about the success or failure of his request. The locally installed mobile phone should be connected with the device through the headphone + microphone socket of the mobile phone. A part of the command will be the password, which is device specific, and is only to be known by the owner of the device. So, any try for illegal access will be denied. Also, the Intelligent Voice Response System (IVRS) will help the user to know whether his request was successful or not. The system disconnects the line automatically if there is no response from the user for a specific amount of time. The user can turn on or turn off any load, and can also control the amount of power delivered to any load using this method. So, this system resembles a remote control system, except only one fact that unlike the IR based or RF based remote controls, our device does not have any distance boundary. You can control this device from any part of the world, where telephone service is available. 2. Using Remote Control Module: The device can be controlled using ordinary TV remote controls too. Most of the TV or DVD-Player remote controls (IR based) can be used with this system. The signal from the remote control is decoded using RC-5 algorithm, which is a popular standard for IR based remote control systems. So, the users don’t have to buy any separate remote control module for this device, most of the TV remotes are compatible with this system. The number buttons on the remote controls can be used for load selection, and volume +/- buttons will control the amount of power delivered to the selected load.