DSP Tools
Processing the Signal in a Different Basis (From Class)
The difficulty in turning a still image into an audio signal is that in order to do so, you must turn a non-time varying signal and turn it into a time-varying audio signal. This was one of the first challenges we were met with and our way of tackling was by changing the basis of our data was in. We now take in RGB values from an image and convert them into a new basis of musical notes. We go through our algorithm, which is expanded upon on the "How We Sonify" page, to convert from RGB values of each section in the random walk into a corresponding musical note.
Image Filtering (From Class)
Part of our algorithm to sonify an image is to determine how many vertical and horizontal lines are in an image and find which of the two is more dominant. If vertical lines are more dominant than horizontal lines then the chords we play will be one octave lower than if there are more horizontal lines than vertical lines. This was added to introduce another level of complexity and give more options to play a piece from the image.
Random Walk (Outside of Class)
One of the benefits of our sonification algorithm is that each time we sonify an image, it will play a randomly different set of chords. This is due to the use of Random Walk in our algorithm, which is a probabilistic method of traversing an image. The algorithm first takes in an image and sections it up into a grid. It then starts in the top left corner of the grid and moves from one section to the next randomly. This method is defined in signal processing as a 2D Random Walk.