1. Musicians and Hearing Aids Marshall Chasin AuD, M.Sc., Reg. CASLPO, Aud(C ) Musicians’ Clinics of Canada

3. Some recent publications…  Canadian Hearing Report, May/June 2007 M. Chasin M. Chasin  Hearing Journal, July 2004 M. Chasin M. Chasin  Trends in Amplification, 7, 3, 2004 M. Chasin and F. Russo M. Chasin and F. Russo  Trends in Amplification, 8, 3 and 4, 2005 King Chung King Chung

4. Special issue on musicians…  Hearing Review March 2006  Edited by Marshall Chasin  www.hearingreview.com (archives) www.hearingreview.com

6. A new book….  Musicians and Hearing Loss (2008) from Plural Publishing.  Replaces my original Musicians and the Prevention of Hearing Loss (1996), Singular Publishing Group.

7. …. Two websites…  www.chasin.ca/music  www.musiciansclinics.com

8. (Five) differences …  Speech vs. Music Spectra  Phonemic vs. phonetic requirements  Differing intensities  Crest factors  (Loudness and intensity)

9. Five differences …  (1) Speech vs. Music Spectra: Speech has a relatively uniform spectrum Speech has a relatively uniform spectrum Human vocal tract sourceHuman vocal tract source Long-term speech spectrum “target”Long-term speech spectrum “target” Music has many sources Music has many sources Highly variableHighly variable No “music target”No “music target”

10. Five differences …  (2) Phonemic vs. phonetic requirements: Speech is mostly low-frequency energy and high frequency clarity (AI). Speech is mostly low-frequency energy and high frequency clarity (AI). Music perceptual requirements depends on the instrument…. Highly variable. Music perceptual requirements depends on the instrument…. Highly variable. Violins need to hear the balance between low and high frequencies.Violins need to hear the balance between low and high frequencies. Clarinets only need to hear the lower frequency energy.Clarinets only need to hear the lower frequency energy.

11. Five differences …  (3) Differing intensities: Speech is 65 dB SPL ± 12 dB Speech is 65 dB SPL ± 12 dB (53 dB SPL to 77 dB SPL)(53 dB SPL to 77 dB SPL) Shouted speech can be 82 dB SPLShouted speech can be 82 dB SPL Music can reach 105 dBA; peaks of 120 dBA Music can reach 105 dBA; peaks of 120 dBA

12. Five differences …  (4) Crest factor: (peak – RMS) Speech has a crest factor of 12 dB Speech has a crest factor of 12 dB Music has a crest factor of up to 18 dB. Music has a crest factor of up to 18 dB. Less damping.Less damping.

13. (Five) differences …  (5) Loudness and intensity: Speech has a well-defined relationship since harmonics are in different critical bands. Speech has a well-defined relationship since harmonics are in different critical bands. Bass instruments have some harmonics within the same critical bandwidth… Bass instruments have some harmonics within the same critical bandwidth… Loudness and intensity mismatch with some instruments (intensity increases and no change in loudness)Loudness and intensity mismatch with some instruments (intensity increases and no change in loudness) Bass instruments “suffer” from this mismatch… underplay…. Less low frequency gain needed…Bass instruments “suffer” from this mismatch… underplay…. Less low frequency gain needed…

14. What about hard of hearing musicians … or non-musicians who like to listen to louder music?

15. Hearing Aids and Musicians 1. Peak input limiting level of most hearing aids limits sound above 85 dB SPL. … great for speech… bad for music. - shouted [a] is about 82 dB SPL peak - music can be >110 dB SPL

18. An Experiment:  Traditional measures of distortion in the industry involve: Autocorrelation Autocorrelation Harmonic relationships Harmonic relationships - total harmonic distortion - intermodulation distortion  Problem: Works best for linear systems Works best for linear systems

19. An Experiment:  Kates (1990, 1991) Notch filter paradigm (can work for non-linear systems) Notch filter paradigm (can work for non-linear systems) Comb filterComb filter With current technology(eg. Adobe Audition), a square notch was removed from the input spectrum With current technology(eg. Adobe Audition), a square notch was removed from the input spectrum “debris” in the notch is a measure of distortion“debris” in the notch is a measure of distortion An output spectrum with a perfectly square notch is representative of perfect fidelity.An output spectrum with a perfectly square notch is representative of perfect fidelity.

20. An Experiment: Unfiltered and Filtered Music

21. An Experiment:  A hearing aid was constructed where the peak input limiting level can be successively reduced from 115 dB SPL, to 105 dB SPL, to 96 dB SPL to 92 dB SPL, … and back to 115 dB SPL.  Acknowledgments: Mead Killion, Russ Tomas, Norm Matzen, Mark Schmidt, Steve Aitken.

22. Unfiltered and Filtered Music

23. Two Input Limiting Levels (115 dB and 105 dB SPL)

24. Two Input Limiting Levels (96 dB and 92 dB SPL)

25. “Signal/Distortion – 30 dB” ratio (0 dB =best)

26. Preference Scales  Gabrielsson et al. (1974; 1991)  Cox and Alexander (1983) Loudness (vs. “faint”) Loudness (vs. “faint”) Fullness (vs. “thin”) Fullness (vs. “thin”) Crispness (vs. “blurred”) Crispness (vs. “blurred”) Naturalness (…“as I remember it”) Naturalness (…“as I remember it”) Overall fidelity (…“dynamics not too narrow”) Overall fidelity (…“dynamics not too narrow”)

27. Signal/Distortion ratio vs. Preference Scale Totals (n=138)

28. Therefore …. Peak Input Limiting Level should be at least 105 dB SPL

29. If you are stuck with a low peak input limiting level ….  Lower volume on stereo or other input and increase gain on aid.  Depending on the manufacturer, you can use an FM system as input.  Use (creative) microphone attenuators (Adhear), or Scotch tape. Tape will provide 7 dB of flat attenuation up to 4000 Hz. (Can also do it electronically- resistive network).

30. Reducing output (not gain)

31. A quick way to determine the peak input limiting level  In a test box, set the stimulus level to 100 dB. (may need to move the reference mic).  Set the gain of the aid to about 5-10 dB, and the OSLP90 is set >110dB. (The output is less than the OSPL-90) (The output is less than the OSPL-90)  If distortion is >10%, then this is “front end” clipping and the peak input limiting level is too low.  Ref: Chasin, 2006, Hearing Journal.

32. Music is intense so…  A useful hearing aid fitting for listening to and playing loud music, even for those with a moderate to severe loss, is a non- occluding BTE fitting.  At higher input levels, less gain and output is required and all that may be needed is 15-20 dB of mid and high frequency gain…

33. A non-occluding BTE mod’n…  Non-occluding BTE provide gain above 1000 Hz and do not occlude the ear canal.  Useful for those with a high frequency loss  BUT still has a front end peak limiting problem…

34. A non-occluding BTE mod’n…  SO…. We can use a desensitized microphone.  Use a high frequency emphasis (-6 dB low frequency roll-off) microphone.  Same frequency response but less front end distortion.

35. Hearing Aids and Musicians 2. One channel is best. - relationship between low frequency fundamental and higher frequency harmonics is maintained (similar CR). - may not be useful in some bass-heavy noisy situations (modification of multi- band?) (L. Revit)

36. Hearing Aids and Musicians … Two channels may be better (?) with one below 500 Hz set to have a higher CR than the higher frequency channel (CR = 1.7:1). This will cut the low frequency gain for all bass sounds.

37. Hearing Aids and Musicians (hearing aids and music) 3. Compression issue. - speech has a crest factor of 12 dB. - music has a crest factor of 18-20 dB. …. Detector should be set 5 dB higher? Depends whether the compression system is based on RMS or peak.

38. WDRC?  WDRC with a higher compression threshold (if it uses a peak detector) may be quite useful…??  Most of damage for mild to moderate loss is OHC… WDRC?

39. Crest factor and output  Music has an 18-20 dB crest factor  Speech has a 12 dB crest factor Therefore the output (OSPL90) of a “music program” in a hearing aid should be 6 dB lower than the program for “speech in quiet”. Therefore the output (OSPL90) of a “music program” in a hearing aid should be 6 dB lower than the program for “speech in quiet”. Also, given similar compression characteristics, the gain should also be 6 dB less intense. Also, given similar compression characteristics, the gain should also be 6 dB less intense.

40. K-AMP or “K-AMP-like” is the best hearing aid for music…  High input limiting level or “equivalent”  One channel  Appropriate compression  WDRC

41. Digital Hearing Aids and Music  Greater dynamic range than 96 dB? (need to be able to transduce inputs >110 dB SPL). Perhaps 20 dB SPL to 116 dB SPL?? 16 bit limitation? 16 bit limitation? A/D converter limitations? A/D converter limitations? … other solutions? … other solutions?

42. Music and a flat response…  I have always “assumed” that a flat frequency response was needed in order to maintain the balance of the low- frequency fundamental energy with the higher frequency harmonic energy….  … This may be secondary…

43. Music and a flat response…  A flat frequency response means a better transient response.  More important for music than for speech. Speech has some transients such as [t]… Speech has some transients such as [t]… Some forms of music have many transients. Some forms of music have many transients.

44. Frequency Transpostion?  Decreasing high-frequencies by about a fifth (eg. G to C) may be useful  Like changing from a violin to a viola…  Examples: Sonovation (non-linear) Widex Inteo Audibility extender (linear) Widex Inteo Audibility extender (linear)

45. Noise and Feedback Reduction  Disable these functions for music??

46. A few slides on feedback reduction techniques…  In general, all approaches use a Fixed slow acting component for static noise Fixed slow acting component for static noise Adaptive fast acting component for dynamic noises. Adaptive fast acting component for dynamic noises.

47. A few slides on feedback reduction techniques…  Three major approaches: (1) adaptive gain reduction (1) adaptive gain reduction eg. Phonak Perseo, Phonak Verve, (Widex Diva/Inteo)eg. Phonak Perseo, Phonak Verve, (Widex Diva/Inteo) (2) adaptive notch filters (2) adaptive notch filters eg. Siemens Triano, Unitron Liaisoneg. Siemens Triano, Unitron Liaison (3) phase cancellation (3) phase cancellation eg. Oticon Syncro/Safron, Resound Canta, Bernafon Symbio, Phonak Savia/Verve, (Widex Diva/Inteo), Unitron Indigo,...eg. Oticon Syncro/Safron, Resound Canta, Bernafon Symbio, Phonak Savia/Verve, (Widex Diva/Inteo), Unitron Indigo,...

48. A few slides on feedback reduction techniques…  Negative issues for each approach: Gain reduction: undesirable gain reduction Gain reduction: undesirable gain reduction Notch filter: Frequency-hopping artifact resulting in a “blurry sound”. Notch filter: Frequency-hopping artifact resulting in a “blurry sound”. Phase cancellation: Chirping- there is no feedback signal for the generated signal to cancel. Phase cancellation: Chirping- there is no feedback signal for the generated signal to cancel.

49. Six feedback reduction techniques and music…  For music…. Since a feedback reduction system may confuse the pure tones of music (and of microwaves, etc) with feedback…  (1) Limit the frequency range that activates the feedback reduction Oticon Syncro and Safron (>1.5 kHz) Oticon Syncro and Safron (>1.5 kHz) Siemens Triano (>2 kHz) Siemens Triano (>2 kHz)

50. Six feedback reduction techniques and music…  (2) Long signal detection (10 sec) with pattern recognition AND a slow adaptive time constant for filter… … short signals are not cancelled. … short signals are not cancelled. (Siemens Acuris/Centra, Bernafon Symbio, Oticon Syncro & Delta, Unitron Indigo, ….) (Siemens Acuris/Centra, Bernafon Symbio, Oticon Syncro & Delta, Unitron Indigo, ….)

51. Six feedback reduction techniques and music…  BUT both (1) and (2) may cause a listener or a player to miss out on the first few bars of music or miss out on the tuning of their instrument.

52. Six feedback reduction techniques and music…  (3) Harmonics in the 3000 Hz region will also have a lower frequency harmonic in the 1500-2000 Hz region, whereas feedback would not (ie., it would be only at 3000 Hz)  Can use a broadly tuned band pass filter from 1000-2500 Hz. If energy, then harmonic; else feedback…..

53. Six feedback reduction techniques and music…  (4) Limit the amount of filter adaptation Create a poorer model of the feedback transmission line so that the feedback reduction technique would be less effective overall. Create a poorer model of the feedback transmission line so that the feedback reduction technique would be less effective overall.

54. Six feedback reduction techniques and music…  (5) Siemens Acuris/Centra. CLAIM: this circuit can distinguish between music and non-music 75-80% of the time. CLAIM: this circuit can distinguish between music and non-music 75-80% of the time. …. It is a two stage phase cancellation approach.…. It is a two stage phase cancellation approach. Distinguishes between internal feedback and an external “feedback-like” tone.Distinguishes between internal feedback and an external “feedback-like” tone. Mitigated by a slow vs. a fast attack time.Mitigated by a slow vs. a fast attack time.

55. Six feedback reduction techniques and music…  (6) Turn off feedback reduction (and noise management) systems.

56. My “gut feeling” regarding music….  If the feedback/noise reduction systems cannot be disabled, the best approach is the adaptive gain reduction since it is the one that is less susceptible to artifacts (eg. Chirping, and frequency-hopping)… (eg. Phonak Perseo, Phonak Verve, Widex Inteo) (eg. Phonak Perseo, Phonak Verve, Widex Inteo)

57. My “gut feeling” regarding music….  … OR … A feed-forward phase cancellation system with SLOW time constants in order to differentiate long lasting feedback from shorter, intermittent musical harmonics. A feed-forward phase cancellation system with SLOW time constants in order to differentiate long lasting feedback from shorter, intermittent musical harmonics. (eg. Bernafon Symbio, Siemens Acuris/Centra, …)(eg. Bernafon Symbio, Siemens Acuris/Centra, …)

58. Some thoughts on noise reduction systems…  DON’T!

59. Some thoughts on noise reduction systems…  Best noise reduction system (other than a binaural hearing aid fitting) is one that uses a directional microphone.  BUT, musicians are so aware of how they are perceived, it is rare to find someone who will wear anything but a SMALL CIC aid.  May be useful for non-musician listeners …

60. Some thoughts on noise reduction systems…  Many manufacturers have tried frequency response changes. This may take the effect of a low cut, a low frequency boost, a high cut, a high frequency boost, or alterations to the mid-frequencies.  None work well, and are probably of minimal use in real life situations.

61. Some thoughts on noise reduction systems…  Modulation rate analysis may offer some hope? The modulation rate is the rate in Hz of the time based waveform increasing and decreasing in intensity. For speech it is typically 4-6 Hz. The modulation rate is the rate in Hz of the time based waveform increasing and decreasing in intensity. For speech it is typically 4-6 Hz. The modulation depth is the amount of decibels between the quietest and loudest elements. For speech this is typically about 30 dB. The modulation depth is the amount of decibels between the quietest and loudest elements. For speech this is typically about 30 dB.

62. Some thoughts on noise reduction systems…  If the modulation rate is greater than 2-3 Hz it is speech or speech-like.  If the modulation rate is less than 2 Hz, it is noise or noise-like.  If the modulation rate is 50 Hz…. ???

63. [a] as in ‘father’

64. Some thoughts on noise reduction systems…  Speech has modulation rates 4-6 Hz  Noise has modulation rates <1 Hz  Music has modulation rates of up to 100 Hz

65. Some thoughts on noise reduction systems…  There are two commonly used types of modulation rate analysis: Multi-channel adaptive modulation rate algorithms. (eg. Siemens, Phonak, Widex, Resound, Oticon, Unitron, Bernafon Neo) Multi-channel adaptive modulation rate algorithms. (eg. Siemens, Phonak, Widex, Resound, Oticon, Unitron, Bernafon Neo) Synchrony detection co-modulation rate algorithms. (eg. Oticon Syncro) Synchrony detection co-modulation rate algorithms. (eg. Oticon Syncro)

66. Some thoughts on noise reduction systems…  Co-modulation rate detection is based on the fundamental frequency (f o ) of speech (100 Hz to 250 Hz).  Noise has no co-modulation rate (ie., it is not periodic).  Music may or may not have a co-modulation rate… so…?

67. A FINAL thought on noise reduction systems… Speech Instrument Music Noise Modulation Rate (Hz) 4-6 Hz 10-100 Hz <2 Hz Modulation Depth (dB) 30 dB 60 dB <15 dB

68. Something else we can do with music and hearing aids…  Many hard of hearing people, especially children cannot enjoy music.  Instead of conditioning hearing aids to accept music, one can change the input.  Many hard of hearing adult musicians have been known to change from violin to viola.

69. Something else we can do with music and hearing aids…  We can transpose music down to a frequency region with better auditory function.  Frequency transposing hearing aids are one solution.  Another is to change the music …

70. Something else we can do with music and hearing aids…  Using software, such as Adobe Audition, music can be transposed without affecting the temporal characteristics.  For example, one can use the “stretch” function while preserving “tempo”.

71. A Plan …  People can bring in pre-purchased CDs with their favorite music.  This can be ripped onto the computer and using Adobe Audition, the frequency of the songs can be reduced/transposed.  They can walk away with the original CD and the new transposed CD, with more intelligible music.

72. www.chasin.ca/music www.chasin.ca/musicandhearingaids www.musiciansclinics.com

73. Musicians and Hearing Aids Marshall.Chasin@rogers.comwww.musiciansclinics.com