Musical Instrument Tones Research Articles

Mismatch negativity to single and multiple pitch-deviant tones in regular and pseudo-random complex tone sequences

Objectives: To determine whether the process responsible for the mismatch negativity (MMN) might be involved in the analysis of temporal sound patterns for information. Methods: Synthesized musical instrument tones of ‘clarinet’ timbre were delivered in a continuous sequence at 16 tones/s, such that there was virtually no N1 potential to each individual tone. The standard sequence comprised 4 or 5 adjacent notes of the diatonic scale, presented either as a regularly repeated, rising pattern or pseudo-randomly. The deviant stimuli were 1–5 consecutive tones of higher pitch than the standards. Results: A MMN was evoked by a single deviant tone, 1 or 5 semitones above the pitch range of the standards. The response to the 5-semitone deviant was significantly larger (mean of 7.3 μV) when the standard pattern was regular as compared with pseudo-random. The MMN latency, on the other hand, was only influenced by the magnitude of pitch deviation. A second MMN was evoked by a second deviant tone, immediately (SOA 62.5 ms) following the first. Further consecutive MMNs were not consistently evoked. Conclusions: The large amplitude of these MMNs can be attributed to the use of complex tones, continuous presentation and a rapid rate of pitch changes, such that no waveform subtraction was required. Over and above the probability with which each individual tone occurs in the standard sequence, the mismatch process is influenced by its temporal structure, i.e. can be regarded as a temporal pattern analyzer. Contrary to the findings of some other groups, we found that two consecutive deviants can evoke an MMN, even at high rates of presentation such that both occur within the postulated ‘temporal window of integration’ of ca. 170 ms. These findings suggest that the mismatch process might be involved in the extraction of sequential information from repetitive and non-repetitive sound patterns.

Auditory evoked potentials to spectro-temporal modulation of complex tones in normal subjects and patients with severe brain injury.

In order to assess higher auditory processing capabilities, long-latency auditory evoked potentials (AEPs) were recorded to synthesized musical instrument tones in 22 post-comatose patients with severe brain injury causing variably attenuated behavioural responsiveness. On the basis of normative studies, three different types of spectro-temporal modulation were employed. When a continuous 'clarinet' tone changes pitch once every few seconds, N1/P2 potentials are evoked at latencies of approximately 90 and 180 ms, respectively. Their distribution in the fronto-central region is consistent with generators in the supratemporal cortex of both hemispheres. When the pitch is modulated at a much faster rate ( approximately 16 changes/s), responses to each change are virtually abolished but potentials with similar distribution are still elicited by changing the timbre (e.g. 'clarinet' to 'oboe') every few seconds. These responses appear to represent the cortical processes concerned with spectral pattern analysis and the grouping of frequency components to form sound 'objects'. Following a period of 16/s oscillation between two pitches, a more anteriorly distributed negativity is evoked on resumption of a steady pitch. Various lines of evidence suggest that this is probably equivalent to the 'mismatch negativity' (MMN), reflecting a pre-perceptual, memory-based process for detection of change in spectro-temporal sound patterns. This method requires no off-line subtraction of AEPs evoked by the onset of a tone, and the MMN is produced rapidly and robustly with considerably larger amplitude (usually >5 microV) than that to discontinuous pure tones. In the brain-injured patients, the presence of AEPs to two or more complex tone stimuli (in the combined assessment of two authors who were 'blind' to the clinical and behavioural data) was significantly associated with the demonstrable possession of discriminative hearing (the ability to respond differentially to verbal commands, in the assessment of a further author who was blind to the AEP findings). Behavioural and electrophysiological findings were in accordance in 18/22 patients, but no AEPs could be recorded in two patients who had clear behavioural evidence of discriminative hearing. The absence of long-latency AEPs should not, therefore, be considered indicative of complete functional deafness. Conversely, AEPs were substantially preserved in two patients without behavioural evidence of discriminative hearing. Although not necessarily indicative of conscious 'awareness', such AEP preservation might help to identify sentient patients who are prevented by severe motor disability from communicating their perception.

Musical instrument identification: A pattern-recognition approach

A statistical pattern-recognition technique was applied to the classification of musical instrument tones within a taxonomic hierarchy. Perceptually salient acoustic features—related to the physical properties of source excitation and resonance structure—were measured from the output of an auditory model (the log-lag correlogram) for 1023 isolated tones over the full pitch ranges of 15 orchestral instruments. The data set included examples from the string (bowed and plucked), woodwind (single, double, and air reed), and brass families. Using 70%/30% splits between training and test data, maximum a posteriori classifiers were constructed based on Gaussian models arrived at through Fisher multiple-discriminant analysis. The classifiers distinguished transient from continuant tones with approximately 99% correct performance. Instrument families were identified with approximately 90% performance, and individual instruments were identified with an overall success rate of approximately 70%. These preliminary analyses compare favorably with human performance on the same task and demonstrate the utility of the hierarchical approach to classification.

Nested modulator and feedback FM matching of instrument tones

Most previous work with nested and feedback frequency modulation (FM) has matched harmonic musical instrument tones with only a single carrier. This paper introduces a genetic algorithm technique for optimizing parameters for several carriers. We give results using the method to match a trumpet, tenor voice and Chinese pipa. The results show that feedback FM often gives the best matches when we compare the techniques against the same number of table lookups, indicating that feedback FM is a good choice for software synthesis. Double- or triple-nested modulator FM gives the best results for the same number of carriers, making them the best choices if nested FM hardware is available.

Tracking the frequency components of musical tones based on global waveform fitting

A novel approach to estimating sinusoidal parameters of musical tones is presented. The proposed algorithm uses a combined quadratic polynomial phase sinusoid and residual signal model, and a waveform fitting criterion, as opposed to the traditional energy fitting one as employed by short-time Fourier transform-based approaches. The proposed algorithm obtains sinusoidal parameters by minimizing the squared difference between the original and the synthesized musical tones. The signal synthesis is implemented implicitly and the synthesis errors are taken into account while the sinusoidal paremeters are estimated. A number of different phase models proposed for music analysis and synthesis are reviewed and their pros and cons are identified. This analysis justifies the use of the proposed signal model. Experiment results for the analysis and synthesis of a variety of musical instrument tones such as piano, trumpet, french horn, flute, and violin are given to demonstrate the accuracy and effectiveness of the proposed algorithm. A VLSI architecture for sinusoid synthesis using the parameters obtained from the proposed algorithm is also given to provide a complete solution to the parametric modeling-based music analysis and synthesis.

Relevant acoustical cues in the identification of Western orchestral instrument tones

The purpose of this research is to establish perceptually relevant operational definitions of the acoustical segments of isolated musical instrument tones. Previous research has yielded conflicting conclusions for continuant (nonpercussive) tones regarding the relative salience of the attack and steady-state portions. This study also examines the perception of impulse (percussive) instrument tones. A new model for the partitioning of continuant tones, the ‘‘amplitude/centroid trajectory,’’ divides the signal into four segments: The attack, attack/steady-state transition, steady state, and decay. The divisions are based on the direction of and relationship among the global rms amplitude and spectral centroid. Other operational definitions of signal partitions and reverse playback were included as editing treatments. Single isolated tones were used from six impulse and six continuant instruments. Nine subjects participated in a 246-item forced-choice identification experiment. It was determined that, for sustained continuant tones, the steady state alone provides sufficient and necessary information for the identification of the instrument. The coevolution of the amplitude and spectral centroid seem important, but the direction does not. A confounding of envelope with identification for staccato tones was found. For impulse tones, both the direction and coevolution of global components are necessary.

Analysis of concurrent timbres with an auditory model

Concurrently sounding timbres (two or more instruments playing on the same or different pitches) sometimes yield unexpected qualities of sound which are exploited in orchestration. A model simulating spectrum analysis by the human auditory periphery is used to explore the differing contributions of individual instruments to the sum sound. Signals are passed through a bank of overlapping bandpass filters whose center frequency spacing corresponds to equal physical distances along the basilar membrane. Thus sinusoidal components do not fall into single ‘‘bins’’ as in a Fourier analysis, but rather spread excitation over a large number of channels. Some partials are resolved in separate channels while other partials are processed in a single channel with others, leading to interactions that determine the overall timbre of the combination. Sampled musical instrument tones are combined into concurrent arrangements and analyzed with the method.

Music analysis/synthesis using the analysis-by-synthesis/overlap-add sinusoidal model

Sinusoidal models have been successfully applied to a variety of speech processing problems, including speech analysis/synthesis. Due to their generality they are also becoming important for the analysis, synthesis, and modification of music signals. The analysis-by-synthesis/overlap-add (ABS/OLA) system is a sinusoidal modeling system which is well adapted to represent and synthesize musical instrument tones. The ABS/OLA system uses an analysis-by-synthesis strategy to determine the parameters of well-resolved sinusoids much more accurately than is possible by direct Fourier analysis, and uses a computationally efficient overlap-add synthesis model that has been generalized to produce artifact-free synthetic tones in the presence of time- and frequency-scale modification. This paper summarizes the ABS/OLA system, details its application to the analysis, synthesis, and modification of musical tones, and presents recent results in its development, including generalization to quasiharmonic tones and improvements in the computational efficiency of analysis.

Statistical study of spectral parameters in musical instrument tones

Methods for short-time spectral analysis [improved over J. Acoust. Soc. Jpn. 33, 233–241 (1977)] and for extracting time-varying amplitude levels and frequencies of partials from musical instrument tones are described. The accuracy of the combined method is examined by applying the method to a synthetic tone having 32 harmonics. Statistical presentations used include histograms and graphs of average sound pressure level spectra, arithmetic mean sound pressure levels, and standard deviations of frequencies and amplitude levels, as well as autocorrelation and cross-correlation functions for various parameters. Seven tones produced by a violin, a flute, an oboe (with and without vibrato), and by a trumpet (nonvibrato) were analyzed. Characteristic features of the time-varying spectral parameters of each musical instrument tone were obtained. Applications of the results to the synthesis of musical instrument tones are considered. As a result, it is suggested that it will be possible to synthesize good synthetic musical instrument tones by realizing the statistical nature of the extracted spectral parameters.

Portamento and glide tone generator having multimode clock circuit

A musical instrument tone generator comprises a clock circuit including means for developing an interval code representing the number of musical intervals to be swept by a tone signal. An output clock signal is produced in response to the interval code for causing the swept tone signal to exhibit a rate of change varying exponentially with time such that a number of musical intervals are swept in a corresponding number of equal time intervals. Means are provided for suitably updating the output clock signal in response to the modification of an on-going frequency sweep in order to maintain the foregoing correspondence between the number of musical intervals swept and the time duration of the sweep. In another embodiment, the output clock signal is effective for operating the tone generator for developing a tone signal sweeping one or more musical intervals in an equal time interval.

Spectral similarities of tones from “especially useful” musical instruments

In a room the ear can track pitch, tone color and dynamic nuances of musical instrument tones with quickness, stability, and precision, even in the presence of many similar tones or of noise (otherwise music would be impossible). It has been recognized that the ear is most effective with tones consisting of a set of harmonic components; with fairly strong lower partials and upper partials that rapidly fall away. This spectral rolloff apparently reduces intercomponent masking in the upper critical bands, a phenomenon that exists above the 6th harmonic. A study has been made of room‐average spectra for tones from instruments recognized as being especially useful in ensemble playing (oboes, violins, trumpets, pianos, saxophones, and bassoons). The spectra are remarkably similar: After a few srong lower harmonics each instrument shows an average rolloff of very nearly 18 dB/octave beginning at the 5th or 6th harmonic. There is no reason in physics why these diverse instruments share such spectral features. Instruments that “carry” less well (e.g., flute, cornet, or “vibes”) roll off earlier and/or more steeply. Tones with slower rolloff (e.g., the lowest note of an oboe sans lower joint) also show weakened trackability and pitch. [Work supported by NSF.]

A method for pitch extraction of musical instrument tones using discrete fourier transform

A method for pitch extraction of musical instrument tones using discrete fourier transform

Perceptual evaluations of synthesized musical instrument tones

An analysis-based synthesis technique for the computer generation of musical instrument tones was perpetually evaluated in terms of the discriminability of 16 original and resynthesized tones taken from a wide class of orchestral instruments having quasiharmonic series. The analysis technique used was the heterodyne filter—which produced a set of intermediate data for additive synthesis, consisting of time-varying amplitude and frequency functions for the set of partials of each tone. Three successive levels of data reduction were applied to this intermediate data, producing types of simplified signals that were also compared with the original and resynthesized tones. The results of this study, in which all combinations of signals were compared, demonstrated the perceptual closeness of the original and directly resynthesized tones. An orderly relationship was found between the form of data reduction incurred by the signals and their relative discriminability, measured by a modified AAAB discrimination procedure. Direct judgments of the relative sizes of the differences between the tones agreed with these results; multidimensional scaling of the latter data provided a visual display of the relationships among the different forms of tones and pointed out the importance of certain small details existing in the attack segments of tones. Of the three forms of simplification attempted with the tones, the most successful was a line-segment approximation to time-varying amplitude and frequency functions for the partials. The pronounced success of this modification strongly suggests that many of the microfluctuations usually found in the analyzed physical attributes of music instrument tones have little perceptual significance.

The Differential Limen of the Attack Time of Pure Tones and the Classification of Various Attack Transient Patterns by a Discriminant Function Technique : On the Attack Transients of Musical Instrument Tones, Part 1

The Differential Limen of the Attack Time of Pure Tones and the Classification of Various Attack Transient Patterns by a Discriminant Function Technique : On the Attack Transients of Musical Instrument Tones, Part 1

Perceptual constancy in complex sound identification.

The aim of the study was to discover whether there are perceptual constancies between classes of complex sounds, even though the acoustic information is very different. Listeners were asked to identify sounds consisting of vowels, musical instrument tones and meaningless buzz sounds, presented at three different fundamental frequencies (octave shifts), and at three signal‐to‐noise differentials. The sounds were numbered but not named. The vowels had voice‐like quality in mid‐octave (110 Hz); the musical tones sounded music‐like at the highest octave (220 Hz); at the lowest octave (55 Hz) both music and vowels were physically meaningless; the meaningless buzz sounds had no greater relevance in one octave than any other. These buzz sounds could be distinguished from each other by differing harmonic patterns. It was found that, in the absence of masking noise, the vowels were better identified in mid‐octave than were the musical and meaningless sounds. The musical sounds were not better identified than the others in high octave. The meaningless sounds were unexpectedly easier to identify in the low octave. They generally withstood thermal noise masking well, but not buzz masking, compared with the musical and vowel sounds. It was more difficult to identify musical and vowel sounds during thermal noise masking than during buzz masking. It would seem the fixed‐ratio hypothesis is not applicable to vowel (formant) sounds within the octaves tested. It may be used, however, to explain the constancy of identification for sounds which depend on harmonic patterning.

Derivative Synthesis of Musical Instrument Tones

Spectral variances, typical for musical instrument tones, occur with changes in loudness level. Nonlinear electronic circuit elements, having appropriately chosen transfer characteristics, can produce spectra, variant with the level of their input signals. Based on this, a system of musical tone synthesis has been developed and a synthesizer was built. Parameters that characterize musical sounds can be programmed into the synthesizer. Such parameters—the major ones being the attack and decay time constants, the nonlinear transfer characteristic, and the formant filter—to be determined through measurements or experimentation. Synthesized sounds of some conventional musical instruments will be demonstrated.

Analysis of Musical Instrument Tones

A technique is presented for the analysis of musical instrument tones. The end result is a mathematical representation for the tone being investigated. Since no assumptions are made about the frequencies of the partial tones, inharmonic frequencies can be detected and accounted for in the representation. An analog-digital converter was used to present the signals to the ILLIAC II digital computer for analysis. Five tones were analyzed, resynthesized, and subjected to listening tests. The trumpet tone was characterized by a frequency change which, if omitted during synthesis, made the tone sound mechanical. The saxophone tone had a short-duration, inharmonic component which gave it the “bite” characteristic of a jazz tone. The other tones analyzed were produced by a violin, bassoon, and clarinet; and each, when resynthesized, still retained the tonal quality of the original; as examples, the wind tones sounded breathy and the violin sounded bowed. It was thus concluded that this analysis technique can be successfully applied to analyzing tones of the classes of instruments represented by these five tones.

Relationships between Wind Instruments Timbre and Sound Spectra

The sounds of nearly all wind instruments are characterized by formant-like patterns that are identified by musicians with descriptive terms such as superior or unacceptable, dark or bright, and the like. This study contributes to the basic understanding of the relationship between the sound spectra of wind musical instrument tones and the terminology used in describing those tones. The basic patterns of all common wind instruments are included. Tape recordings will be used to demonstrate the relationships presented.