How To, Education & Training

Development of the Adaptive Music Perception Test

Description
Development of the Adaptive Music Perception Test Martin J. Kirchberger, 1,2 and Frank A. Russo 3,4 Objectives: Despite vast amount of research examining the influence of hearing loss on speech perception,
Published
of 11
7
Published
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Similar Documents
Share
Transcript
Development of the Adaptive Music Perception Test Martin J. Kirchberger, 1,2 and Frank A. Russo 3,4 Objectives: Despite vast amount of research examining the influence of hearing loss on speech perception, comparatively little is known about its influence on music perception. No standardized test exists to quantify music perception of hearing-impaired (HI) persons in a clinically practical manner. This study presents the Adaptive Music Perception (AMP) test as a tool to assess important aspects of music perception with hearing loss. Design: A computer-driven test was developed to determine the discrimination thresholds of 10 low-level physical dimensions (e.g., duration, level) in the context of perceptual judgments about musical dimensions: meter, harmony, melody, and timbre. In the meter test, the listener is asked to judge whether a tone sequence is duple or triple in meter. The harmony test requires that the listener make judgments about the stability of the chord sequences. In the melody test, the listener must judge whether a comparison melody is the same as a standard melody when presented in transposition and in the context of a chordal accompaniment that serves as a mask. The timbre test requires that the listener determines which of two comparison tones is different in timbre from a standard tone (ABX design). Twenty-one HI participants and 19 normalhearing (NH) participants were recruited to carry out the music tests. Participants were tested twice on separate occasions to evaluate test retest reliability. Results: The HI group had significantly higher discrimination thresholds than the NH group in 7 of the 10 low-level physical dimensions: frequency discrimination in the meter test, dissonance and intonation perception in the harmony test, melody-to-chord ratio for both melody types in the melody test, and the perception of brightness and spectral irregularity in the timbre test. Small but significant improvement between test and retest was observed in three dimensions: frequency discrimination (meter test), dissonance (harmony test), and attack length (timbre test). All other dimensions did not show a session effect. Test retest reliability was poor ( 0.6) for spectral irregularity (timbre test); acceptable ( 0.6) for pitch and duration (meter test), dissonance and intonation (harmony test), melody-to-chord ratio I and II (melody test); and excellent ( 0.8) for level (meter test) and attack (timbre test). Conclusion: The AMP test revealed differences in a wide range of music perceptual abilities between NH and HI listeners. The recognition of meter was more difficult for HI listeners when the listening task was based on frequency discrimination. The HI group was less sensitive to changes in harmony and had more difficulties with distinguishing melodies in a background of music. In addition, the thresholds to discriminate timbre were significantly higher for the HI group in brightness and spectral irregularity dimensions. The AMP test can be used as a research tool to further investigate music perception with hearing aids and compare the benefit of different music processing strategies for the HI listener. Future testing will involve larger samples with the inclusion of hearing aided conditions allowing for the establishment of norms so that the test might be appropriate for use in clinical practice. Key Words: Adaptive test, Hearing impairment, Hearing loss, Music perception. (Ear & Hearing 2014;XX;00 00) 1 Department of Health Sciences and Technology, ETH Zürich, Zurich, Switzerland; 2 Phonak AG, Stäfa, Switzerland; 3 Department of Psychology, Ryerson University, Toronto, Ontario, Canada; and 4 Communication Team, Toronto Rehabilitation Institute, Toronto, Ontario, Canada. INTRODUCTION Motivation Hearing impairment affects the perception of sound leading to reductions in speech intelligibility and the enjoyment of music. Extensive research into the effects of hearing loss on speech intelligibility (e.g., Humes 1991; Barreto & Ortiz 2008) has led to various attempts to improve speech perception through hearing aids (Levitt 2001; Tawfik et al. 2009; Cornelis & Moonen 2012). By comparison, research into the effects of hearing loss on music perception is in its infancy. Hearing impaired (HI) listeners unaided and aided have specified various ways in which the enjoyment of music is affected: instruments are hard to distinguish, sounds are distorted, and melodies are difficult to recognize (Neukomm 2006; Sarchett 2006; Leek et al. 2008). There are no standardized tests available that have been designed to assess music perception in individuals with hearing loss adaptively. A large number of music perception tests are specifically designed to assess musicality or the skills of healthy and formally trained individuals (Woodruff 1983; Don et al. 1999). Another group of music tests focuses on individuals with cochlear implants (Gfeller et al. 2002; Gfeller et al. 2005; Medel Medical Electronics 2006; Cooper et al. 2008; Looi et al. 2008; Spitzer et al. 2008; Kang et al. 2009). The first music perception test battery to be specifically developed for hearing-aid users was the Music Perception Test (Uys & van Dijk 2011). This test battery includes 11 subtests that assess perception in four dimensions: rhythm (4 subtests), timbre (2 subtests), pitch (2 subtests), and melody (3 subtests). As the tests are not adaptive, its sensitivity to differences between NH and HI listeners could be somewhat limited. Concept Music can be broken down into a core set of components such as meter, harmony, melody, and timbre. Meter is the division of a rhythmic pattern according to equal periods or measures (Limb 2006). Strong and weak beats in the pattern are grouped into larger units (Oh 2008), typically categorized as duple or triple according to whether the measure is organized in groups of two or three beats. The perception of harmony corresponds to the vertical organization of pitch (Piston 1987). This organization concerns simultaneous pitches as in a chord or in counterpoint. Another important aspect of harmony concerns the sequencing of chords, as in a harmonic progression. A melody is a linear succession of tones, which is perceived as a single entity (Limb 2006). The succession of tones can be specified with regard to melodic contour and interval size (Davies & Jennings 1977). The former refers to the direction of the pitch change (Schubert 2006) and the latter refers to its magnitude. Timbre is often defined as the attribute of auditory perception whereby a listener can judge two sounds as dissimilar using any criterion other than pitch, loudness, and duration (ANSI 1960). This definition is somewhat unsatisfactory in that it defines timbre through what it is not 0196/0202/14/XXXX-0000/0 Ear & Hearing Copyright 2014 by Lippincott Williams & Wilkins Printed in the U.S.A. 1 zdoi; /AUD 2 KIRCHBERGER AND RUSSO / EAR & HEARING, VOL. XX, NO. X, XXX XXX rather than what it is (Sethares 2005). An alternative approach to defining timbre uses multidimensional scaling techniques to extract the underlying physical dimensions that contribute to its perception (Grey 1977; Krumhansl 1989; Krimphoff et al. 1994; McAdams et al. 1995). Brightness and attack are two dimensions that have been consistently identified as important using multidimensional scaling techniques (Grey 1977; Wessel 1979; Krumhansl 1989; McAdams & Cunibile 1992; Krimphoff et al. 1994; McAdams et al. 1995). The conclusions regarding a third dimension, however, differ. Within the framework of our study, we have defined a third dimension in accordance with the conclusions of Krimphoff et al. (1994) as spectral irregularity. The spectral irregularity is determined by deviation from linearity in the amplitude envelope of the harmonics. A clarinet would have high spectral irregularity because it lacks energy in the even harmonics. A trumpet tone by contrast would have low spectral irregularity as it possesses odd and even harmonics with a linear spectral roll-off. Objective We aimed to develop a computerized test that is adaptive and thus capable of assessing music perception in hearing as well as HI populations. The Adaptive Music Perception (AMP) test has been designed to investigate the perception of important music dimensions such as meter, timbre, harmony, and melody. These mid-level dimensions are constructed from 10 low-level dimensions D1 to D10 such as attack, level, and pitch (Fig. 1), which may ultimately be modified in a fairly direct manner through changes to the hearing aid technology. While the assessment is not entirely culturally neutral, there is no reason to expect that performance will vary across individuals who primarily consume music in the Western harmonic idiom regardless of genre preferences. Each test of the battery asks the listener to make a response concerning a mid-level dimension while adaptively manipulating the underlying low-level dimensions, thereby allowing for the determination of the respective low-level discrimination thresholds. Thus, thresholds are determined within a musically relevant framework. We contend that the use of these tests corresponds more directly to functional hearing in music than a conventional assessment of discrimination thresholds. Hypothesis Survey studies of individuals with hearing loss have revealed a wide range of problems encountered following the onset of hearing loss including melody recognition and level changes (Feldman & Kumpf 1988; Leek et al. 2008). We assume that these subjective impressions of problems can be objectively tracked with adaptive listening tests that are presented within a musical framework. We predict that individuals with hearing loss will have higher discrimination thresholds than individuals with NH. MATERIALS AND METHODS The AMP test comprises four subtests including meter, harmony, melody, and timbre discrimination. General Test Design Each subtest was designed to determine the discrimination thresholds of two or more low-level physical dimensions. In each trial, one of these low-level dimensions was manipulated D1: Level Meter Harmony Melody Timbre D2: Pitch D3: Duration D4: Dissonance D5: Intonation MUSIC Fig. 1. The mid level and low-level dimensions of the Adaptive Music Perception test. D1 D10 indicates dimension 1 10; MCR, melody-to-chord ratio; Sp. Irreg., spectral irregularity. so as to instantiate the mid level dimension. For example, the downbeat in the meter subtest could be conveyed by a difference in level, duration, or pitch. The adaptive process of each low-level dimension was interleaved to obscure the structure of the test and to discourage analytic modes of listening. We pseudorandomized the position of the correct answer with the constraint that the right answer was equally distributed among both answer buttons. Adaptation An experimental procedure is adaptive if the selection of the stimuli is determined by the results of the preceding trials (Treutwein 1995). Our motivation for making the AMP test adaptive was to measure performance in a manner that was fast and accurate for all degrees of hearing loss and music ability. The AMP test applies the transformed two-alternative-forcedchoice method using a two-down one-up rule (Levitt 1971), which is commonly used in tests of auditory thresholds (e.g., Jesteadt et al. 1977; Wier 1977; Moore et al. 1984; Bochner et al. 1988). The presentation and adaptation of the low-level dimensions in each subtest were interleaved. The adaptation process was stopped, once a minimum of six reversal points were reached on each dimension. The standard step sizes were halved after the first and second reversals in an effort to accelerate the determination of thresholds without compromising accuracy. Discrimination thresholds were calculated by averaging the values of the last four reversals. Sound Stimuli Stimuli used across all four subtests were digitally synthesized. A spectrogram of the stimuli is displayed in Figure 2. The reference tone that was used for calibration and throughout D6: MCR I D7: MCR II D8: Brightness D9: Attack D10: Sp. Irreg. KIRCHBERGER AND RUSSO / EAR & HEARING, VOL. XX, NO. X, XXX XXX 3 the tests was a complex tone with the fundamental frequency f 0 = 220 Hz (pitch A3) and 40 harmonics. The starting phases of the 40 harmonics were randomized once and equal for all participants. The amplitude of each harmonic ( n = 0, 1, 39) was attenuated by 2 db n in relation to the amplitude of the fundamental f 0. The resulting spectral shape corresponds roughly with that of a trumpet tone (Krimphoff et al. 1994). The test stimuli were played back at 40 db SL which is consistent with previous psychoacoustic studies (Jesteadt et al. 1977; Wier 1977; Moore et al. 1984; Bochner et al. 1988; Arehart 1994). Test Procedure The AMP test was implemented in MATLAB R2009a. Before testing with a participant commenced, the hearing thresholds were determined for the reference tone using a twoalternative forced-choice procedure. The presentation level of the reference tone in the test was then individually set for each participant at 40 db SL. On average, the meter and timbre subtests took 10 min each; the harmony subtest took 12 min; and the melody subtest took 11 min. Each subtest could be interrupted and continued at any time. Each subtest was preceded by training trials with feedback to ensure that the task was understood. The initial discrimination differences of the low-level dimensions were made high enough so that the nature of the tasks did not generate confusion. To assess test retest reliability for each subtest, participants were asked to complete testing on two separate sessions. Meter Subtest: Level, Pitch, and Duration A sequence of six harmonic tones was played isochronously in each trial. The participant was asked to indicate whether the sequence constituted a duple or triple meter. The interonset interval between isochronous beats was 600 msec. The upbeats had the spectral characteristics of the calibration tone (fundamental frequency: 220 Hz, roll-off: 2 db per harmonic) and a duration of 280 msec. As mentioned above, the initial phases are defined and originate from a one-time randomization before the test development. The downbeats are realized by adaptively introducing a difference from the upbeats with respect to level, pitch, or duration. Correspondingly, the dimensions for which thresholds are determined are labeled level (D1), pitch (D2), and duration (D3). Rationale for Stimulus Design If the interonset interval is lower than 100 msec, listeners perceive a sequence of tones as a single, continuous event. For intervals greater than 1. 5 sec, grouping gets more difficult as the sounds seem disconnected (Fraisse 1978). In between these two limits, the distance of two events is overestimated for small intervals and underestimated for bigger intervals. The point of subjective equality, where subjective judgments corresponded to the objective duration, is reached at approximately 600 msec (Krumhansl 2000). The sensation of pulse is also most salient for intervals of 600 msec (Parncutt 1994). To increase the sensitivity of the meter subtest, we therefore chose to use an interonset interval of 600 msec. The perceived meter of a sequence is generally established quite early after a couple of beats (Parncutt 1994). With regard to the test design, it is beneficial to keep the test duration short to avoid fatigue. In the meter subtest, only six beats are played back in each trial. This is the least common multiple of duple (two) and triple (three) meter and allows for three or two full measures respectively. According to participants feedback in the pilot study, six beats are sufficient to detect the meter if the downbeat is audible. Harmony Subtest: Dissonance and Intonation Two cadences composed of three-dyad chords (root note and fifth) were played back in succession: subdominant (IV), dominant (V), and tonic (I) (Fig. 3). The participants were asked to indicate the cadence that appeared to be less resolved. The reference cadence was perfectly in tune, the second cadence was detuned. We used just intonation to define the frequencies of the reference cadence: the ratios of two notes in any interval were defined by the harmonic series and are related by small whole numbers. The frequency ratio of the root note and the fifth of every dyad was 3: 2, such that the respective harmonics Fig. 2. Spectrogram of the Adaptive Music Perception test audio stimuli. The reference tone (0 0.4 sec) in the calibration, meter subtest, and timbre subtest is followed by the reference cadence of the harmony subtest ( sec), and by one example of the melody subtest ( sec). IV V I Fig. 3. The subdominant (IV), dominant (V), and tonic (I) dyad of the harmony subtest. 4 KIRCHBERGER AND RUSSO / EAR & HEARING, VOL. XX, NO. X, XXX XXX coincide. Had we used equal temperament, the frequency ratio 12 of the root note and the fifth would be ( 2) 7 = , and thus the harmonics would not coincide (Fig. 4). The frequency ratio between the subdominant and the tonic was 4: 3. The root note of every dyad was composed of 6 harmonics with a roll-off factor of 2 db per harmonic. The fifth of every dyad was composed of 4 harmonics with a roll-off factor of 3 db per harmonic. The first harmonic of the fifth was attenuated by 1 db compared to the first harmonic of the root note. By this means, the coinciding harmonics (root: harmonic 3/fifth: harmonic 2, root: harmonic 6/fifth: harmonic 4) were equal in amplitude. The initial phases of all harmonics were pseudorandomized with the provision that coinciding harmonics differed by 120. In one of the two cadences, the tonic dyad was manipulated in one of two different ways: either the pitch of the fifth was increased or the pitch of the whole dyad was shifted. In the first case, the vertical structure of the tonic dyad was detuned causing higher sensory dissonance (Terhardt 1984; Tufts et al. 2005). In the second case, the vertical structure of the tonic dyad was in tune; however, the presented pitch did not correspond to the expected pitch in the given tonality (Piston 1987; Krumhansl 1990; Lerdahl 2001). The adaptive modifications that were applied to the pitch of the last dyad were within the bounds of a semitone. Thus, the basic meaning of the cadence was preserved (Blackwood 1985) and the pitch modifications implied a change in intonation. Accordingly, the dimensions for which thresholds were determined in the harmony subtest were labeled dissonance (D4) and intonation (D5). Melody Subtest: Melody-to-Chord Ratio Two 4-note melodies were played back and accompanied by a harmonic chord, which served as a mask. The task was to identify whether the contours of the two melodies were the same or different. The two melodies presented during the trial always started on different pitches (i.e., second melody was transposed). The transposition (pitch shift by constant interval) of the melodies was implemented to ensure that focusing on absolute pitches could not circumvent the melody recognition task. If the contours of the two melodies were the same, the second melody would be a copy of the first melody chromatically transposed by one whole tone. The contour of either the first or the second melody was made different by interchanging the second and the third tone of the respective melody. Two sets of melodies, a lower and a higher one, were included in the subtest. The two sets as well as the accompanying A-major and G-major chords are depicted in Figure 5. During the subtest, the sound level of the chords was kept constant and the levels of the melodies were adapted. For both melodies, the melody-to-chord ratio (MCRs) were determined indicating the point at which the melodies were perceptually masked by the chords. The dimensions are labeled MCR I (D6) and MCR II (D7) correspondingly. Timbre Subtest: Brightness, Atta
Search
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x