Introduction

When deciding on what to focus on to create a project around, I knew that whatever the project might be it should revolve around audio. I first decided that it should be an audio processor that could generate some audio effects. After playing with a vocoder one day I thought that it would be good idea to try to build my own so that I could recreate that classic robotic sound. So I decided to build a vocoder so as to build a system that takes a voiced signal and modulates it with a harmonic signal, thus creating that classic robot sound.

After doing some research I decided I wanted to build a more modern vocoder that reduces the signals bandwidth yet retains more of the signals properties. The original vocoder, the Channel Vocoder, for instance reduces the bandwidth yet can greatly distort the original sound. The phase vocoder is one newer encoding scheme that seemed to be mentioned in numerous audio papers for having good uses in audio processing. I assumed that all the vocoder types are similar in their effects on the signal and that they differ primarily in their implementation and bandwidth/bit rate reductions. Later, I found this assumption to be horribly wrong. How they split up the signal is very different and they all use different methods to model speech.

So the project evolved into the following. Create a Phase Vocoder for audio applications and attempt to recreate that classic robot voice.

History

The name 'Vocoder' is a contraction of 'Voice' and 'Encoder'. It was conceived by H. Dudley in the 1930's to reduce the amount of bandwidth that speech occupies when being transmitted. It successfully reduced the bandwidth by separating the vocal tract information from the excitation and transmitting this information at a reduced bit rate. A series of bandpass filters split up the signal into bands of frequencies, this along with the voicing information and pitch information are used to reconstruct the signal at the receiving end. It was successfully demonstrated at the 1939 New York Fair and later used by electronic musicians to create some interesting effects.

From research and interest in the channel vocoder, new types of voice encoders began arising. More reasearch began looking at how the speech is produced so as to model the speech for reduced bandwidth in transmission. Four main types of vocoders arose. They are Channel, Homomorphic, Formant and Phase.