We have created a few songs based on a spoken sample of Robert Frost reading his poem, “Nothing gold can stay.” To do this, we analyzed the resonance pattern of a recording of Frost reciting the poem, capturing the frequency and amplitude of the oral, nasal and fricative formants throughout the utterance; and, we derived the sine-wave replicas of Frost’s speech on the lowest three oral resonances, the nasal and fricative resonances and the intermittent bursts and aspirations that give his speech distinctiveness. 

There are several sound files here illustrating sine-wave synthesis. First, we present a comparison of Frost’s natural version and a sine-wave replica that exhibits the resonance pattern faithfully, within the tolerance of our estimate.

To the right of the poet, we show spectrograms of the first line of the Natural and the Sine-Wave versions.

Next, we offer several sine-wave Frost variations, in which the frequency values of the time-varying tones that replicate Frost’s utterances have been forced to express frequencies restricted to the well tempered scale. 

The algorithm for revising the natural formant estimates musically required us to choose specific notes of the scale to use; and, required us to set a single fixed duration for each note throughout an utterance. Within each chunk of time, the average value of a tone was calculated and reset to match the nearest note that we had licensed. 

The effect of setting different durations for the notes can be heard in these examples that use the white notes of the piano keyboard. At one extreme with notes of 50 ms the individual pitches are barely discernible. At the other extreme with notes of 350 ms the sine-wave pattern differs so greatly in temporal grain from the natural resonance pattern that intelligibility is compromised.

Last, the effect of different note specifications can be heard in three examples of the complete poem, each using a nominal note duration of 150 ms. The variations are: chromatic (white keys and black keys), diatonic (white keys only), and pentatonic (black keys only). The spectrograms illustrate the different tone patterns of the first line of the poem.

Acknowledgments: Kathy Dubowski and Judith Meer performed the initial acoustic analyses of the poet’s voice which became the sine-wave synthesis parameters; Kate Simon created the digital hurdy gurdy which was used to force the sine-wave synthesis parameters to exhibit musical durations and pitches; and, Daria Ferro and Robert Remez performed the perceptual assessment of the musical and linguistic qualities of the text setting.  We are grateful to NIDCD (000308) for supporting the lab during the development of this project. 

To cite this page, you may use the URL, and:

Remez, R. E., Ferro, D. F., Simon, K., Dubowski, K. R., & Meer, J. (2007). Musical and poetic sine-wave speech. Technical Report, Speech Perception Laboratory, Barnard College, New York, NY.

Speech Perception Laboratory

Barnard College, Columbia University