Function

The function wave2duration reads a wave file, and outputs the corresponding time sequence.

Parameters

There are two parameters in this function, as follows:

• frameSize: the size of a frame
• overlap: the overlap between neighboring frames

Experiments 1

I change the value of frameSize and overlap individually, to see the impact of them on the result.

I do tests on three wave files: test.wav (133KB, 6.1 seconds), happybirthday.wav (1,585KB, 18.3 seconds), and ChineseBirth.wav (25,520KB, 148.1 seconds). Both 'test.wav' and 'happybirthday.wav' have only one channel. But there are two channels (for left ear and right ear) in 'ChineseBirth.wav'. In the function, I choose one channel from it to do the test.

Experiment 1.1: Fix the value of overlap and change the value of frameSize

Experiment 1.2: Fix the value of frameSize and change the value of overlap

Experiments 2

I plot the Root Mean Square of different versions of the song Happy Birthday to You, to see whether they are similar.

Follows are the versions I use:

• Happy Birthday to You sung by the same person, at different time: rivet_hp1.wav, rivet_hp2.wav, rivet_hp3.wav, rivet_hp4.wav.
• Happy Birthday to You sung by my friend (a male): rodin_hp1.wav is the one he sang when I told him the title of the song; rodin_hp2.wav is the one he sang after he heard what I sang and tried to imitate me.
• Happy Birthday to You sung by another friend (a female this time): cj_hp1.wav is the one she sang when I told her the title of the song; cj_hp2.wav is the one she sang after she heard what I sang and tried to imitate me.

The result is here:

The clustering hierarchy of these waves:

Back to Li Wei's Homepage