Maureen Mellody

The time-frequency characteristics of violin vibrato: Modal distribution analysis and synthesis

A high-resolution time-frequency distribution, the modal distribution, is applied to the study of violin vibrato. The analysis indicates that the frequency modulation induced by the motion of the stopped finger on the string is accompanied by a significant amplitude variation in each partial of that note. Amplitude and frequency estimates for each partial are extracted from the modal distribution of ten pitches that span the range of the violin instrument. The frequency modulation is well-represented by a single sinusoid with a mean rate of 5.9 Hz and a mean excursion of +/- 15.2 cents. A spectral decomposition of the amplitude envelopes of the partials shows that the peaks lie primarily at integer multiples of the vibrato rate. These amplitude and frequency estimates are used in an additive synthesis model to generate synthetic replicates of violin vibrato. Simple approximations to these estimates are created, and synthesized sounds using these are evaluated perceptually by seven subjects using discrimination, nonmetric multidimensional scaling (MDS), and sound quality scoring tasks. It is found that the absence of frequency modulation has little effect on the perceptual response to violin vibrato, while the absence of amplitude modulation causes marked changes in both sound quality and MDS results. Low-order spectral decompositions of the amplitude and frequency estimates also occupy the same perceptual space as the original recording for a subset of the pitches studied.

Analysis of Vowels in Sung Queries for a Music Information Retrieval System

A method for analyzing and categorizing the vowels of a sung query is described and analyzed. This query system uses a combination of spectral analysis and parametric clustering techniques to divide a single query into different vowel regions. The method is applied separately to each query, so no training or repeated measures are necessary. The vowel regions are then transformed into strings and string search methods are used to compare the results from various songs. We apply this method to a small pilot study consisting of 40 sung queries from each of 7 songs. Approximately 60% of the queries are correctly identified with their corresponding song, using only the vowel stream as the identifier.

Vowel representation and synthesis of singer identity

Professional singers shape the acoustical attributes of their vocal production to create the perceptual impression of a single instrument spanning a variety of vowel timbres over a wide range of pitch. Samples of the five Italian vowels over a two‐octave range were analyzed using the modal distribution to yield a time‐frequency image of the signal. From such images (high‐order) representations of the composite transfer function of the signal were extracted. These representations provide little evidence for the existence of an acoustic ‘‘signature’’ that is invariant over the singer’s entire range of production. Rather, the data, along with results from a perceptual study, suggest that singers create the impression of a single instrument by smoothly transitioning from one local region of invariance to the next. By identifying these regions, low‐order, perceptually robust approximations are constructed which bear the identity of the singer when used to synthesize new instances of sung vowels. These low‐orde...

Modal distribution analysis of vibrato in musical signals

Due to the nonstationary nature of vibrator notes, standard Fourier analysis techniques may not sufficiently characterize the partials of notes undergoing vibrato. Our study employs the modal distribution, a bilinear time-frequency representation, to analyze vibrato signals. Instantaneous frequency and amplitude values for each partial are extracted using Hilbert techniques applied to local neighborhoods of the time-frequency surface. We consider vibrato in violin and vocal performance. Our study confirms the presence of both amplitude modulation and frequency modulation in the partials of notes generated by each of these instruments, and provides a fine-grained analysis of these variations. In addition, we show that these instantaneous amplitude and frequency estimates can be incorporated into methods for synthesizing signals that perceptually resemble the original sampled sounds.

Monaural phase effects in the discrimination of the spectral shape of transients

Limits on temporal resolution in auditory perception range from 25 ms, for the judgment of temporal order, to a lower limit of 3–4 ms for the discrimination of monaural phase. Results are presented on the discrimination of spectral shape for signals concentrated within this lower bound. A wideband, 4‐ms noise was compared with spectrally smoothed versions. The phase spectrum was controlled by assigning the same random phase spectrum to both the original and smoothed signals. Depending on the choice of phase spectrum, discrimination thresholds for spectral shape were found to vary from 2–3 dB to as much as 15–17 dB, at which point the smoothed spectrum is essentially flat. This dependence on phase can be eliminated by presenting a train of transients, rather than a single transient. Such findings do not appear to be accounted for by synthesis artifact, nor by any simple features of either the complex spectrum or waveform envelope of these wideband signals. Phase effects are observed, however, in a computat...

Applications of time‐frequency analysis in musical acoustics

The modal distribution (MD) is a member of Cohen’s class of time‐frequency distributions that is designed specifically for signals that are well‐modeled as the sum of time‐varying partials, such as those generated by many musical instruments [W. Pielemeier and G. H. Wakefield, J. Acoust. Soc. Am. 99, 2382–2396 (1996)]. When combined with local operators on the time‐frequency surface, MD analysis provides substantial improvement over techniques based on the spectrogram with respect to simultaneously resolving variations in amplitude and frequency. Furthermore, the analysis degrades gracefully as the acoustic signal varies from the specified model and, in several cases, can be extended to handle a broader class of signals. Examples drawn from violin vibrato, singing, and the piano are used to illustrate MD analysis and its extensions. Issues of system identification, time‐varying signal models, and musical synthesis are also discussed on the basis of these examples. [Research supported by grants from the Ford Motor Company, the Office of Naval Research, the National Science Foundation, and the Office of the Vice President for Research at the University of Michigan.]