Using a wah-wah pedal guitar is something guitar players have to learn. Recently, more intuitive ways to control such effect have been proposed. In this direction, the Wahwactor [1] system controls a wah-wah transformation in real-time using the guitar player's voice, more precisely, using the performer [wa-wa] utterances. To come up with this system, different vocal features derived from spectral analysis have been studied as candidates for being used as control parameters.

Although the wah-wah effect was initially developed by trumpet players using mutes in the early days of jazz, it has became known as a guitar effect ever since Jimi Hendrix popularized Vox Cry-baby pedal in the late 60's, see Figure 1.

A wah-wah guitar pedal contains a resonant bandpass filter with a variable center frequency that is changed by moving the pedal back and forth with your foot. Usually, a knob controls the mix between original and filtered guitar signals as represented in Figure 2.

The explanation of the why wah-wah effect resembles human [wa-wa] utterance is found on the voice spectral characteristics of the vocalic phonemes [u] and [a], in particular, on the first formant location. Considering the [u] vowel first formant is around 350 Hz, and the [a] vowel first formant is around 700 Hz [2], the [u] to [a] articulation produces a modulated sound due to the trajectory of the first formant that is perceived as the effect of a resonant filter moving upwards in frequency.

There is already a musical interface developed by ATR Media Integration & Communication Research Labs that profits from the link between the wah-wah effect and the [wa-wa] utterance. The system, called Mouthesizer [3], uses a video camera to measure the opening of the performer's mouth and changes the wah-wah filter centre frequency according to this measure. It was in fact the multifaceted requirements of such a system what made us think about an alternative straightforward solution.

The Wahwactor is a two-input and one-output system. Out of the two input tracks, one of the tracks may be considered a control rather than a proper input since it is the one in charge of driving the transformations to be applied to the audio signal. In the context of the Wahwactor, the audio signal is typically a guitar signal and the control signal is a voice [wa-wa] utterance signal.

First, the voice signal is analyzed to pick up a meaningful descriptor that, after a simple conversion (shift, scale and smooth), is used as the centre frequency of the wah-wah filter, through which the guitar signal is sent to be mixed with the original.

To work in real-time, the Wahwactor uses a frame-by-frame algorithm described by the diagram illustrated in Figure 4. The voice signal is sampled at 44100 Hz and analyzed using a 2100 sample Hamming window and a 2048 point Fast Fourier Transform. The guitar signal is sampled at 44100 Hz and filtered using 2100 sample length buffers. The algorithm uses a 1050 sample hop size so that we have a 50% overlap in synthesis. This overlap is necessary to smooth the filter phase frame to frame variations.

We have been able to test the Wahwactor with several guitarists in a set of informal sessions. The conclusion we can derive from such evaluations is that controlling a wah-wah effect by uttering [wa-wa] is meaningful, intuitive, and very easy to do with no previous knowledge of 'playing with a wah-wah pedal' required. We believe this is mainly due to the fact it takes much more time and practice to learn how to tap with your feet than to learn how to tap using your voice, not only physically but also because of psychomotor considerations.

Apart from the wah-wah, other effects have been implemented including filter based effects such as phaser or flanger, and envelope based effects such as tremolo. None of them has proven to neither be as intuitive or show a relevant phonetic-sound linkage as in the wah-wah.

Further work includes reducing the latency (which actually is approximately 10 ms for sound I/O using ASIO drivers and 24 ms due to the analysis hop size), taking care of background noise to make it more robust in a real live music performance environment, and applying different mappings at different pitches.

- The paper by Alex Loscos & Thomas Aussenac, "The Wahwactor: a voice-controlled pedal", Proceedings NIME 05, International Conference on New Interfaces for Musical Expression, May 26-28 2005, Vancouver, Canada.

- A photograph of our brilliant poster, presented at NIME 05.

- Two video demos:

1.Guitarist playing chords with the Wahwactor	2.Guitarist playing a riff with the Wahwactor

- The Wahwactor VST plug-in v0.2 made following the Virtual Studio Technology SDK and CLAM (C++ Library for Audio and Music). So far, this plug-in is for Windows XP only (if someone is motivated to code it for Mac OS X let us know...). You can use it with hosts like CubaseSX and Plogue Bidule (that we highly recommand). DO NOT HESITATE TO SEND US ANY FEEDBACK.

This project has been partially funded by the EU-FP6-IST-507913 project SemanticHIFI.

REFERENCES

CONTACT

LINKS