1

I am trying to find a software that can compare several waveforms together and extract the similarities between them.

Basically, I have a few pieces which have been generatively composed and I wish to see how similar they are to each other in order to be able to see whether there is some kind of rhythmic, frequency, or volume, recurrences that could help me understand how different the same piece is at different instances.

I want to be able to say 'Oh hey, at around 3mins I always get this particular density of sounds at this frequency," or "oh hey, there seems to be a similar volume intensity at these three different points although the sound changes."

Hope someone out there can help me out, maybe I'm just using the wrong keywords in my search... Cheers everyone!

  • 2
    Note that two waveforms can look almost completely different, yet sound essentially identical from a musical or human perspective. Are you sure you want to compare waveforms, and not, at least, spectrograms? – Linuxios Jul 17 '17 at 16:25
  • Yeah probably a spectrogram would make more sense, but I would still need to have a horizontal view so to speak of peaks and throughs. And anyways, it depends on whatever software I find.. the problem is I do not know what to look for. – Istishhad Hheva Jul 29 '17 at 12:10
  • @Linuxios I'll be honest, I've always seen them as the same thing. What's the difference? – Hashim Aug 9 at 4:23
  • @Hashim: You've always seen what as the same thing? Waveform and spectogram? – Linuxios Aug 9 at 5:32
  • 1
    @Hashim: In the sense that a signal and its Fourier transform contain the same information, you're absolutely right. But by ignoring the phase component of the frequency domain information, we're able to consider the content of the signal in a way that much more closely represents the mechanisms of human hearing. (Unless I've misunderstood your question...) – Linuxios Aug 9 at 5:38
2

Essentia is an

Open-source library and tools for audio and music analysis, description and synthesis

It includes some pre-built extractors that you can use.

0

Maybe you could use an eq that has a "matching" option (like izotope ozone eq) to compare the frequency content directly but I dont think there's anything more you can compare.

0

You might want to take a look at the Voxengo Span Plus plugin. It at least allows you to display the spectrum from different channels/tracks simultaneously (at varying frequency resolution), but I think it may also let you compare different spectrums, http://www.voxengo.com/product/spanplus/features/

0

I highly recommend studying the spectrum of these sounds. You can identify all the frequencies in a portion of sound, and how they change over time and their amplitudes. This is evident with a 2D or 3D spectrogram. The frequencies involved are essentially defining the timbre of the sounds.

You can download iZotope RX for viewing but also repairing and processing the spectral content. You can also view the audio waveforms alongside. For real-time analysis there is also iZotope Insight, which is also excellent. These are specialised tools, so you can adjust the resolutions to suit the analysis.

Aside from finding the phase of certain frequencies (which is generally difficult to hear), the spectrum is going to be most useful. It is not easy to 'hear' phasing when just observing a limited length of audio either.

You can also display a spectrum with the freeware software, Audacity.

0

The freeware Sound-Similar Free is the right tool.

Sound-Similar (Free Version) is a lightweight software application that measures the perceptual similarity between two WAV files in Linear PCM format, the most commonly used format in WAV files. The degree of similarity is represented by a similarity score in percentage, from 0% to 100%. The software does not compare the two digital files bit by bit. Nor does it measure the resemblance of the two waveforms. The similarity is rated based on human auditory perception through advanced time, frequency, and time-frequency domain analyses. The resultant similarity score can be used for sound classification as well as perception-based sound quality check.

The two WAV files to be compared can have different sampling rates, bit depths, and one or two channels of data. If the file is stereo, then average values of the two channels will be used for comparison. Sound volume difference does not affect the similarity measurement unless it is too low to maintain the perceptible sound quality.

The frequency ranges of different classes of sounds, such as human speeches, music and environmental sounds, may be different. Sound-Similar allows the user to specify the comparison frequency range in order to increase the scoring accuracy. It is possible to extend this range to infrasonic or ultrasonic region.

There are two comparison modes: (1) Full Length vs Full Length, and (2) Full Length of the Shorter vs Partial Length of the Longer. In both modes, the two WAV files can have different lengths in time. Mode 2 can be used to check whether the shorter is a part of the longer and, if YES, where the shorter is located in the longer. The similarity scoring algorithm in the software has been optimized for general-purpose use. A similarity score below a few percent usually indicates that the two sounds are totally different, while a score from a few percent to 100% often shows that the two are alike with a varying degree of similarity.

It can be downloaded at: http://www.sound-similar.com

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.