ReSound: An Audio Tool Increasing the Resonant Dimension of Recorded Sounds

The aim driving this research was to find an alternative method of audio editing that allowed for unwanted sound events to be cleaned from a sound file without distorting the tonal integrity of its’dynamic peaks. Journal articles were found that identified separation as the solution to this problem, where the dynamic signal would be separated from the transient noise information. This would allow for goal-oriented editing to be applied directly to the source of contention. The program, ReSound, was created as a proof of this concept. Once the separation was accomplished, two types of modulation processes were introduced into the signal and noise sections. The noise information had Frequency Modulation applied so that bands of frequencies not containing the unwanted sound events could be identified and increased in gain. Amplitude Modulation was applied to the signal section so that the frequency bins containing the dynamic peaks could be isolated and raised in gain. After programming the modulation processes, a synthesis method was researched that would closely follow the evolution of a sound file’s envelope over time. The Audio Signal Driven Synthesis method was applied to the modulation processes so that an unaffected carrier signal could be indirectly modified by the modulated copy of the carrier signal. Testing this method in the ReSound program found that Audio Signal Driven Synthesis maintained the harmonic presence introduced by the initial sound file. Additional processing effects were incorporated into the program providing the user with various enriching ways to edit the final output.

Hello my name is Nicolette Cash and today I will be presenting my senior research project.

Background:

Introduction:
I went outdoors and recorded the sound in a heavily wooded area. I was able to collect really nice details like birds chirps and rushing water, but when I listened to my recording at home I heard very car drive by. I sought to reduce that unwanted sound event by editing my sound in a digital audio workspace. After adding two effects, I came across the problem that the dynamic bird chirps were very flat when compared to how they sounded in the initial recording.

Question:
From this problem rose the question, how can a sound be edited so that it is as clear as the original location recording?

Research Findings:
Research found that separating a sound file’s dynamic signal from its’noise floor would allow for an unwanted sound event to be filtered from the noise transients while preserving the clarity of the sound file’s dynamic peaks [8].

Concept:
These findings developed the concept that to increase the clarity of recorded sounds during the editing process, a program could be engineered that separated the sound file into two variables of signal and noise. Those of which would be individually modulated by amplitude of the signal and frequency of the noise. These two modulations would then be summed into a resynthesized output that preserved the recordings’harmonic frequencies and allowed for additional processing effects to enrich the final output.

Research:
The workflow of ReSound functioned as a proof of this concept.

File I/O:
A recorded sample of natural sounds was introduced into the Max/MSP program. The source file is opened, and the sonogram shows the frequency magnitudes of the unedited sound file.

Separation:
The ReSound program separates the recording’s signal from the noise using a spectral subpatcher that analyzes the frequency information from the source file.
The user adjusts the noise to signal ratio to only include the noise information so that they can set a value at which all of the noise information is contained.

Noise:
After the separation is complete, the user is then allowed to edit a range of frequency bands for the Noise section to increase frequencies that don’t include the unwanted sound event.

Signal:
In the signal section, the user can reset the ratio slider of noise to signal to only include the signal information and then alter the gain, or volume, sliders to modulate the signal’s amplitudes. The aim of this area is to increase the frequency sliders that hold the dynamic peaks like bird chirps.

Resynthesis:
The program then outputs a resynthesized signal to noise floor ratio based on the user’s noise floor value, and a sonogram shows the frequency magnitudes after the separation process.

Effects:
Further processing effects are added based on the user’s needs. Overdrive increases the sharpness of the chosen section, cascade increases the final output’s color based on dramatic filter shapes, and teeth alters the resonant complexity of the final output based on the delay parameter settings.

Finally the output can be recorded and saved to a user determined location [4].

Conclusions:
Testing the research findings during the programming process of ReSound identified “Frequency, Amplitude, and Spectral Envelopes as key elements for reconstructing resonant audio signals similar to their initial source environment.

Visualizer Comparisons:
It was concluded that affecting a sound’s resonance required boosting or reducing the partials of a sound. The more partials that were adjusted the better because this essentially tailored the sound’s fingerprint. The sonogram comparison showed that separation of the noise and signal information did maintain the dynamic peaks while reducing the unwanted sound events.

Improvements:
Future Modifications:
An input with a particularly noisy background could mask the dynamic subtleties of the signal section so a signal inverter could be programmed into ReSound offering the user another approach to cancel out any offending sound events.

Thank you for watching my presentation. I hope you enjoyed. If you have any questions or comments please feel free to submit them, and I will provide an answer as soon as possible.