Pitch Modulation Function

Main Objective

The purpose of the pitch modulation function is to enable the DJ to modify the soundtrack by creating a variation in sound in the music, in real time. This can be done by shifting the pitch by a number of semitones or using vibrato to create a rapid change in pitch. Through pitch modulation, the music is able to effectively evoke more emotion and expression, which adds on to the ambience of a live performance. The user has the option to use any WAV file of their choice for pitch modulation. 

Function Algorithm

A phase vocoder is used to manipulate phase information in order to alter the pitch. 

First, the short-time Fourier transform (STFT) is applied to the signal in order to access its frequency components. To apply the STFT, the signal is broken up into smaller segments of a specified length, called windowed segments. A windowing function, such as the rectangular or Hamming window, can be applied in order to do this. The windowed segments each have a certain overlap, called hop length (defines the length in between windows). This overlap of windowed segments is to prevent information loss between frames. 

The length of the windowed signal as well as the hop length are hard-coded and not set by the user. The values 1024 and 256 were used for the length of the windowed signals and the hop length, respectively. A larger window size would improve frequency resolution while sacrificing time resolution. Typically, a window size of 2048 would be used for balanced time and frequency resolution; however, we use a slightly smaller window size to account for possible rapid changes in the audio signal. While a smaller hop length would improve temporal smoothness, a larger hop length would reduce overlap, which may cause "choppiness" in the signal.

Pitch Modulation Effect on Signal Waveform

Below are the signal waveforms, before and after pitch modulation. We can see that increasing the modulation frequency will create a faster oscillating signal, as shown by the rapid variation near the amplitude of the signal. We can see that increasing the modulation depth will increase the fluctuation of the magnitude of the signal, as the maximum amplitude is increased.

Demonstration

DSP Tools Utilized in the Pitch Modulation Function

• Short-time Fourier Transform: This is the Fourier transform performed on small segments of signal. These segments are created by a windowing function that is slid across the signal with certain overlap (called hop length). Further details on how this is applied can be found in the above section, Function Algorithm.

Results and Challenges

Overall, the pitch modulation function was successfully able to modify audio in real time. A phase vocoder was used to enable pitch modulation by manipulating phase information. By changing the modulation frequency and modulation depth, we were able to achieve a pitch modulated output using this function. However, the audio output ended up having extra noise and static, mainly due to impulses in the audio (such as bass). To ensure robust audio output, we can apply a low-pass filter to filter out the noise of the modulated output before writing it to the output stream. This would ensure that the modulated output is minimally affected and would be more promising for use in a DJing environment.

One of the biggest challenges was learning how to use the Python libraries on top of learning Python, as each library had many available functions to choose from. Another challenge was learning how to make the GUI and have real time audio manipulation, which was implemented through the use of various libraries such as tkinter and threading.

Images adapted from:

https://sethares.engr.wisc.edu/vocoders/phasevocoder.html (Image 1)

https://course.ece.cmu.edu/~ece491/lectures/L25/STFT_Notes_ADSP.pdf (Images 2 and 3)