Applied Signal Processing

Projects, to be handed in before February 15.


  1. Neural Networks


In many cases we have a problem where we cannot easily formulate an algorithmic solution, but we can get lots of examples of the behaviour we require. In such a case a neural network might work. For example, our brain is a neural network. In chapter 26 of Smith (http://www.dspguide.com/ch26.htm) an introduction is given about neural networks, and some examples are given. Here the aim is to write a neural network to recognise numbers on license plates.


  1. See http://prettyview.com/ann/nnfs.shtml for an implementation of this problem. Now write a program that can recognise numbers on license plates.

  2. Train the program using the dataset on http://prettyview.com/ann/nnfs.shtml. If needed, use more datasets.

  3. Now run the program on model license plates, to which you have added noise, and give an estimate of the successrate of your program.

  4. Make some photographs of some real license plates, and apply your program to them. Discuss the result.

Another introduction on neural networks can be found at http://www.doc.ic.ac.uk/~nd/surprise_96/journal/vol4/cs11/report.html#Introduction%20to%20neural%20networks