pitch discrimination of tunned percussion instruments for cochlear implants users

Upload: federico-nahuel-cacavelos

Post on 23-Feb-2018

213 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/24/2019 Pitch Discrimination of Tunned Percussion Instruments for Cochlear Implants Users

    1/5

    1

    Acoustic Laboratory December 2015, Argentina

    PITCH DISCRIMINATION OF TUNED PERCUSSION INSTRUMENTS

    FOR COCHLEAR IMPLANTS USERSCacavelos, Federico Nahuel

    Ingeniera de Sonido, Universidad Nacional de Tres de [email protected]

    AbstractThe present study focuses the analysis on the pitch discrimination of impulsive sounds. This study aims to

    compare experimentally the competences of normal hearing (NH) and cochlear implant users (CI) by stimulatingwith tuned percussion instruments present in rhythmic patterns. The hypothesis of the study is impulsive signals

    should be more complicated for discrimination than continuous note because the temporal fine structure. For this

    reason different percussion instrument are tuned in an easily recognizable pitch for CI users and presented in

    rhythmic patterns. ABX test was carried out in 21 NH subjects and 1 CI at the moment. The results are compared

    with previous studies where the test signal was continuous tone. The result showed that in NH subjects the

    discrimination rate is preserved. More CI is needed to test the hypothesis with large interval confidence.

    1. INTRODUCTION

    The research on music perception of Cochlear

    Implant users (CI) has been developing during the

    last years. Most of the studies investigate the abilityof melodic contour identification and timbre

    recognition, as well as emotional communication,

    rhythm and meter recognition. However, there are no

    currently many studies addressing the perception of

    impulsive signal sounds and possible identification

    CI users. Therefore this study investigates the

    identification of impulsive signals by using

    percussion instruments that were tuned in toneswhere CI should easily distinguish.

    In the current section are shown some results of

    previous studies. Methods and procedures aredescribed in the second section. Results of the test

    will be shown in third section and compared with

    previous studies in the fourth section. The fifth

    section is reserved to conclusions.

    The central auditory processing in human been has

    two possible mechanism to perceive pitch. One ofthem is based on the temporal theory, where the

    perception of time periods smaller than 1 ms enabling

    us to perceive frequencies up to 1 kHz. This is

    essential to recognize the fine structure of signals. InCI this mechanism can be hampered by the low

    stimulation rates used in some codification strategies.The other mechanism used by the central auditory

    processing is the place theory, where describes how

    resonances distributed in the basilar membrane

    produce a frequency dispersal, commonly called the

    tonotopic distribution [1]. In CI is not possible to

    stimulate the cochlea in two sectors simultaneously

    due to electrical interference between electrodes, this

    way the codification strategies keep just the peaks

    signal of a period, discarding the week intensity. This

    is why CI users have difficult to detect fine structure

    of pitch information and impulsive sounds.

    Cognitive factors may also agree difficult to

    identification process. Disturbance Auditory

    Processing is a hearing impairment where exists an

    impediment to analyze and / or interpret soundpatterns. In this sense is necessary to attend the

    central auditory processing. [2]; [3]

    It is well established that people with hearing

    impairments, including CI users, perceive rhythm

    approximately as well as those with normal hearing[4], [5], [6], and [1].

    Gfeller et Lansing (1991) [7] administered a testcalled PMMA (Primary Measures of Music Audition

    developed for Gordons in 1979) to 18 postlingually

    deafened CI subjects. Mean scores on the rhythm

    subtest (88%) were higher than on the tonal subtest(78%).

    Leal et al. (2003) [8] conducted rhythm

    discrimination tasks in twenty-nine postlingually

    deafened adults scoring 95%.

    Kong et al (2004) [9] found no significant difference

    in performance among four tempo condition (60, 80,

    100, and 120 beats per minute) in normal hearing and

    cochlear implant users by an identification task where

    asked to subjects to read and chose the musicalnotation displayed on the screen that corresponded to

    the rhythmic pattern presented.The collective findings across a range of studies

    indicate that CI users perform significantly worse

    than NH controls on pitch-based tasks.

    In the 1991 Gfeller study [7], CI subjects scored 78%

    on the tonal subtest of the PMMA, by varying the

    second note in four semitones, while NH scoredhigher than 95%.

    Schulz & Kerber (1994) [13] assessed pitch

    perception for eight users of a single-channel CI

    programmed with a now-obsolete analogue

    processing scheme. In the melodic-directionperception task, subjects were required to assess

  • 7/24/2019 Pitch Discrimination of Tunned Percussion Instruments for Cochlear Implants Users

    2/5

    2

    whether a tonal sequence played on the piano was

    ascending, descending, or unchanging in pitch. Theresults of this assessment indicated that, subjects with

    CIs scored between 68% and 84%, while NH scored

    100% suggesting that although they could undertake

    the task with varying degrees of accuracy and ease,

    their overall scores were still significantly below thatof the NH control group. Also they asked todifferentiate whether two tones played on a piano

    were the same or different in pitch. The results

    indicated that whilst NH subjects could reliably

    differentiate between notes one semitone apart, CI

    group were able to differentiate intervals larger than a

    major second or a minor third. The authors also

    reported that these results were fairly consistent

    across the 11 octave frequency range used in this

    task.

    Fujita & Ito (1999) [10] found that five of eight CI

    users discriminated the higher of two notes between

    four and ten semitones apart. However, the remainingthree subjects could not discriminate between two

    notes, one octave (i.e. 12 semitones) apart.

    Gfeller et al. (2002) [15] also investigated pitch

    perception by comparing CI users with NH controls

    in a higher/lower pitch discrimination task. Whereasthe mean difference limen for NH subjects was 1.13

    semitones the CI populations performance was

    highly variable with a mean difference of 7.56

    semitones.

    Eunoak Kim (2012) [5] also tested the CI capabilities

    by using a synthesized piano tone in fourth basedfrequencies (185 Hz, 262 Hz, 330 Hz and 390 Hz).

    The normal hearing group scored 96% (SD, 3%)correct while the CI listeners scored 70% (SD, 11%)

    correct. These fourth based frequencies are used in

    the current study for tune the different sound

    material. This way can be compare both test results.

    A recent study compared pitch discrimination ability

    at 1 and 6 semitones using real-world stimuli of

    sung vowels sounds. CI subjects were severely

    impaired compared to NH subjects. NH performed

    89% and CI 60% for the 6 semitone changecondition. On the 1 semitone change condition NH

    perform on 81% and CI 49% of accuracy (Sucher and

    McDermott, 2007) [4].

    Phillips-Silver et al (2015) [11] test the capacity offeeling beat of drum music in CI users by playing

    different latin merengue rhythms and analyze thebody movements by a accelerometer remote control.

    They found that cochlear implant users are able to

    synchronize dance-like body movement to drum

    music (even as well as they entrain to visual and

    auditory metronomes). In comparison with a group of

    matched controls, however, the synchronizationperformance of CI users is worse across all

    conditions. Also found evidence that pitch variations

    would interfere with beat finding and

    synchronization.

    The objective of the present study is to evaluate the

    pitch identification skills of percussive instrumentssounds that have strong impulsive information when

    they are played in a simple rhythmic pattern. The

    hypothesis is that CI users should perform worse on

    pitch discrimination of impulsive signals (percussion

    instrument) than in tone signals due to the finestructure.The method of analysis that this study propose is

    using the same parameters discrimination pitch tone

    signals from previous studies and thus compared with

    the results founded of impulsive pitch sounds signals.

    For this reason a three different tuned sounds was

    made tanking into account enough harmonic

    components difference being easily recognizable

    even for CI users.

    There are certain limitations of the study since some

    parameters that cannot be controlled. The loudness

    variation will depend strongly of the calibration

    processor settings and physiological condition. Theuser can detect and discriminate the sounds only by

    the loudness difference. Also cognitive aspects of

    memory loss and attention may hamper the test,

    where the difficulty of the discrimination does not

    imply a hearing problem but rather a deficiency in thecentral auditory processing [2], [3].

    Some previous studies explain that exist wide

    variation on the perception skills in CI users [4], [5],

    [6], [1], it can attach disadvantages when analyzing

    the confidence intervals of the data collected. This is

    because differences in the CI devices, codificationstrategies and the patient own adaptation.

    2. PROCEDURE

    2.1 Stimuli PreparationSource signal consist in rhythmic patterns made with

    three percussion instruments as Kick, Snare and

    Cymbal. Each percussion instrument was carefully

    processed and transpose to the same frequency of

    tone signals from previous studies.

    Figure 1: The three pitched instrument spectrums used inthe patterns.

    This way, each instrument has 6 semitones distance

    from each other, F3 (170 Hz) for the kick, C4 (261

    Hz) for snare and G4 (390 Hz) for ride cymbal, as is

    possible to see by their spectrums on figure 1. This

    distance was intended to bring clearly pitchidentification according to Eunoak Kim (2012) study

  • 7/24/2019 Pitch Discrimination of Tunned Percussion Instruments for Cochlear Implants Users

    3/5

    3

    [5]. The three sounds instrument was normalized in

    peak amplitude because of peak method detectionused in the common coding strategies.

    The three percussion instruments were used to create

    different rhythmic patterns used as test material.

    The test is divided in two parts with propose to

    achieve more specify of the subjective identification.In the first part the same instrument is used to thewhole rhythmic pattern, and compares the three

    instruments separated in each pattern. An example of

    the MIDI sheets to the first part is shown in Figure 2.

    Figure 2: The three signals of rhythmic patterns used forfirst part. Top: Ride cymbal in the whole pattern; Middle:

    Snare in the whole pattern; Bottom: Kick in the whole

    pattern.

    The second study combines the three instruments in

    each pattern and changes just one hit as it is possible

    to see in figure 3. The first part attempts to discern if

    they really perceive the difference in impulsive pitch

    and the second aims to analyze discrimination in a

    combined pattern of different instruments. This offers

    more complex to understanding of the phenomena.

    The number of hits for all rhythmic configuration was

    6, showing good appreciation by Kim in 2012 [5]

    after testing with different quantities of tones.

    Figure 3: The three signals of rhythmic patterns used for

    second part. Top: Ride cymbal hit change in combinedpattern; Middle: Snare hit change in combined pattern;

    Bottom: Kick hit change in combined pattern.

    The 15 combinations involve the repetition of tests on

    the same subject in order to increase the number of

    trials and improve the confidence interval. Since it is

    expected more variation in the results of the second

    part, the first part was repeated two times whilesecond part was repeated three times. For each

    repetition a different pattern was used confronting the

    subject always to different signals.

    All sounds where made in Albeton Live 7.0 using an

    Electric Drum Roland DR-909 and DR-707 andtranspose of drum rack of native software. The tempo

    condition was 120 bpm following the results of Ying-

    Yee Kong (2004) [14] where it does not find real

    difference in de discrimination skills when the beatsper minute change.

    2.2 Subjective Test

    ABX test with 15 stimuli combinations was made

    repeating two times the first part of test and threetimes the second part, each time the patterns waschange getting different type of stimuliavoiding thesubject to facilitate remembering the previous pattern

    choice. The subjects has to identify if the last signal

    (X signal) is equal to A (X=A) or if it is equal to B

    (X=B).

    The test was applied at the moment in 21 NH subjects

    of 20-55 years old and 1 bilateral postlingual CI user

    of 52 years old (3 years of experience with CI)

    equipped with Nucleus 5 (22 electrodes and CP810

    processor using ACE strategy). A Matlab algorithm

    was used in order to apply order randomization of the

    audio presentation. Test environment was differentbetween both groups of study. In the CI user the line

    out of the processor was used to avoid the

    characteristics of the acoustic environment, but in NH

    subjects a headphone Audio Technica ATH-D49 was

    used into a quiet room with acoustic treatment. In allcases a Notebook computer with Maudio FastTrack

    was used to reproducing the audio test. Just one ear

    was stimulated in all cases. Total time of test was 5

    minutes in each subjectdepending on how long youuse each subject to choose the answer. To replay the

    audio was not allowed but the subject can choose hisown conformably sound level.

    There are some parameters that cannot be controlledin both groups. The most important factor is the

    degree of attention that the subject can achieve at that

    moment. This can be hampered by limited memory

    capacity and slower processing speed since disorders

    of the central auditory processing according to Oscar

    Caete (2006) [3]. Also different models, settings and

    codification strategy of the CI devices can add wide

    variation to the identification process of the subject.Several authors also point out that there is greatvariation between the perceptual skills of users [6],

    [1]. This happens because of the resilience of the

    subjects to electrical stimulation of their particular

    physiological characteristics.

    3. RESULTS

    The percentage of hits was calculated for all subjects

    using Matlab software. In order to get the differences

    between different instruments, the results are

    presented by each comparison. The three first groupsof columns correspond to the first test and second

    group of columns correspond to the second test.

    By the assumption that the statically analysis of an

    ABX test follows a binomial distribution, according

    to J. Boley in 2009 [12]. It is possible to obtain by a95% of confident interval of each result.

  • 7/24/2019 Pitch Discrimination of Tunned Percussion Instruments for Cochlear Implants Users

    4/5

    4

    As is shown in the figure 4 and table 1, the normal

    hearing identified nearly 100% in all the stimulipresented, showing just lower performance for the

    second part of the test. From the three signals

    presented in the second part, which corresponds to

    the change of snare-ride has a slight increase of the

    percentage of successes. This result is consistent withthe frequency of the tuning given distance.

    Figure 4: Bar graphic of percentage of successes for

    Normal Hearing subjects (NH) and Cochlear Implant users(CI), for first part of test (left columns) and second part

    (right columns) with a 95% confident interval (black lines).

    Table 1: Values of percentage of successes for Normal

    Hearing subjects (NH) and Cochlear Implant users (CI) and

    the 95% confident intervals.

    May be notice that for the CI results it is not possible

    to get a confidence interval useful to analyze the

    results globally.

    4. DISCUSSION

    Normal hearing subjects perform better than 89 %

    correct in the combined patterns and almost 100% forthe patterns with the same instrument. These results

    can confirm some of the previous studies:

    89% (Sucher CM, 2007) [4]

    95% (Gfeller K, 1991) [7]

    100% (Schulz and Kerber, 1994] [13]

    Although, the numbers of trials to CI are insufficient

    to permit a real data analysis, is possible to see that

    CI user performed worse than any NH performance.

    Results of the study evaluated to date in the only

    cochlear implant user reveals lower performance than

    previous studies:

    60% (Sucher CM, 2007) [4]

    78% (Gfeller K, 1991) [7]

    68% - 84% (Schulz and Kerber , 1994) [13]

    Is necessary to point that the previous studies use a

    continuous note and this study is using an impulsive

    sound, in this sense is expected lower performance

    for impulsive sounds.Some errors with NH group may have been caused

    by the subject poor attention. Although a drop in

    performance was not shown with over time in both

    groups.

    The subject evaluated with CI showed some difficult

    to some discrimination task, and the forced choice

    was obliged to respond randomly. At the same way

    he expresses that he was unable to identify certain

    sounds but he can recognize the difference by not

    because their pitch components but rather for their

    loudness. One of the reasons of this event is because

    CI users may have different performance of their

    calibration. Sounds can match with specific channelshaving different impedance and thus generate

    loudness differences, easy recognizable for the CI

    users.

    5. CONCLUTIONS

    The hypothesis cannot be tested because the large

    confident intervals of the CI users. This can be

    improved by increasing the number of trials in CI

    users, achieving more confidence results. This studyis currently in progress.

    Possible future studies can include others factors by

    using more frequencies in order to achieve greater

    accuracy in the cochlear implant identification,

    although previous studies of continuous tone showed

    no variation in frequency scale.Some cochlear implant users can identify a difference

    between samples but does not mean that they

    perceive correctly the sound. For this reason, future

    works may consider most number of instrument

    sounds with different harmonic composition can

    provide better specificity to the phenomenon

    6.

    REFERENCES[1] Valeriewei Lool, Music Perception of Cochlear

    Users, Doctorate Thesis, Department ofOtolaryngology , University of Melbourne, (2006).

    [2] I.J.R. Restrepo, J.R.C. Medina, Desrdenes del

    procesamiento auditivo,IATREIA, 19(4), (2006).[3] Oscar Caete S., Central Auditory Processing

    Disorder, Revista Otorrinolaringologica Cabeza

    Cuello, 66, 263-273, (2006);

    [4] Sucher CM and McDermott HJ., Pitch ranking of

    complex tones by normally hearing subjects and

    cochlear implant users, Hear Res, 230, 80-87, (2007).

    [5] Eunoak Kim et al., Music Perception Ability of

    Korean Adult Cochlear Implant Listeners, Clinical

    Kick-Snare

    (SAME)

    Snare-Ride

    (SAME)

    Kick-Ride

    (SAME)

    Kick-Snare

    (COMBINED)

    Snare-Ride

    (COMBINED)

    Kick-Ride

    (COMBINED)

    NH 100% 95% 98% 90% 89% 94%

    NH 95%

    Confident

    Interval

    0% 6% 6% 9% 9% 7%

    CI 100% 50% 100% 66% 58% 50%

    CI 95%

    Confident

    Interval

    0% 49% 0% 19% 39% 40%

  • 7/24/2019 Pitch Discrimination of Tunned Percussion Instruments for Cochlear Implants Users

    5/5

    5

    and Experimental Otorhinolaryngology, 5(1), S53-

    S58, (2012).[6] P. Donnelly and C. Limb Music, Perception in

    Cochlear Implant Users. The Johns Hopkins

    University, Baltimore, Maryland (2009).

    [7] Gfeller K and Lansing CR, Melodic, Rhythmic,

    and Timbral Perception of Adult Cochlear ImplantUsers, Journal of Speech, Language and HearingResearch, 34, 916-920 , (1991).

    [8] Leal M.C. et al., Music Perception in Adult

    Cochlear Implant Recipients, Acta Oto-

    Laryngologica, 123, 826-835, (2003).

    [9] Kong YY et al., Music perception with temporal

    cues in acoustic and electric hearing, Ear Hear, 25,

    173185, (2004).

    [10] Fujita S et al., Ability of nucleus cochlear

    implantees to recognize music, Annals Otol

    Rhinology Laryngology, 108, 634-40, (1999).

    [11] J. Phillips-Silver, Cochlear implant users move

    in time to the beat of drum music, Elsevier, (25-23).Canada (2015).

    [12] J. Boley, Statistical Analysis of ABX Results

    Using Signal Detection Theory. Audio Engineering

    Society, 127, 7826, (2009).

    [13] Schultz E, Kerber M., Music perception with theMED-EL implants, Advances in cochlear implants,

    326332,(1994).

    [14] Ying-Yee Kong et al., Music Perception with

    Temporal Cues in Acoustic and Electric Hearing,

    Lippincott Williams & Wilkins. Ear & Hearing,

    25(2), (2004)

    [15] Gfeller K et al., Recognition of familiar

    melodies by adult cochlear implant recipients andnormal-hearing adults, Cochlear Implants, 3(29-53),

    326332, (2002).