Literature Overview - Digital Hearing Aids and Group Delay - HADF, June 2017, P. Derleth
Historic Context Delay in HI and the perceptual effects became a topic with the widespread market introduction of digital hearing aids in the 90 s Digital HI typically have delay s between 2-10ms (in special operation modes up to 13ms, e.g. excange of audio-signals between L&R device) Analog HI had typically around ~1ms of delay which did not trigger investigations about potentially negative perceptual effects Research on DAF (delayed auditory feedback) in the 50 s focused on delay effects (30-300ms) on speech production Research on the effects of delay between Audio and Visual information in the 80 s found negative effects in the delay range 40-160ms P. Derleth, May 2017 Page 2
Literature Stone M.A. & Moore B.C.J. 1999. Tolerable hearing aid delays. I. Estimation of limits imposed by the auditory path alone using simulated hearing losses. Ear Hear, 20, 182 192. Agnew J. & Thornton J.M. 2000. Just noticeable and objectionable group delays in digital hearing aids. J Am Acad Audiol, 11, 330 336. Stone M.A. & Moore B.C.J. 2002. Tolerable hearing aid delays. II. Estimation of limits imposed during speech production. Ear Hear, 23, 325 338. Groth J. & Søndergaard M.B. 2003. Disturbance caused by varying pro pagation delay in nonoccluding hearing aid fi ttings. Int J Audiol, 43, 594 599. Stone M.A. & Moore B.C.J. 2003. Tolerable hearing aid delays. III. Effects on speech production and perception of across-frequency variation in delay. Ear Hear, 24, 175 183. Müsch H., Nooraei N., Edwards B. & Pavlovic C. 2004. Permissible delay in a partially occluding monaurally fi t hearing aid. Poster presented at IHCON 2004, LakeTahoe, USA. Stone M.A. & Moore B.C.J. 2005. Tolerable hearing-aid delays IV. Effects on subjective disturbance during speech production by hearing-impaired subjects. Ear Hear, 26, 225 235. Stone M.A., Moore B.C.J., Meisenbacher K. & Derleth R.P. 2008. Tolerable hearing aid delays. V. Estimation of limits for open canal fi ttings. Ear Hear, 29, 601 617. Lars Bramsløw (2010) Preferred signal path delay and highpass cut-off in open fittings, International Journal of Audiology, 49:9, 634-644 Continous but sparse scientific publication activities P. Derleth, May 2017 Page 3
Delay in HI - Definition Audio Processing Delay (measured in s) in a HI-System - Phase Delay: for sinusoidal signals: phase delay ~ phase change / frequency - Group/Envelope/Wave-packet Delay: for signals consisting of multiple sinusoidal components: a) phase delay is identical for all components => envelope preserved (shifted in time), i.e. linear-phase-system typical for block-processing, i.e. FFT based HI b) phase delay is frequency dependent => envelope distortion typical for sample-processing, i.e. TD based HI, filter-bank P. Derleth, May 2017 Page 4
Delay in HI Envelope Figures θ Δt ~ Δθ / f A f Δt t θ A env. distort. ~ dθ / df f Δt t P. Derleth, May 2017 Page 5
Delay in HI - System View Mech. -> Elec. Micophone 0.1ms TD Processing 1-2ms Elec. -> Mech Receiver 0.2ms Mech. -> Elec. Micophone 0.1ms A -> D Conversion 0.25ms FFT Processing 5-7ms TD Processing 2-5ms D -> A Conversion 0.25ms Elec. -> Mech Receiver 0.2ms P. Derleth, May 2017 Page 6
Delay in HI - System View Binaural Audio Mech. -> Elec. Micophone 0.1ms A -> D Conversion 0.25ms wait for contralateral Signal ~5ms FFT Processing 5-7ms TD Processing 2-5ms D -> A Conversion 0.25ms Elec. -> Mech Receiver 0.2ms P. Derleth, May 2017 Page 7
Delay in HI Multi Audio Path External sound sources: - direct air acoustic path - delayed (HI processed) acoustic path Own voice: - bone acoustic path - direct air acoustic path - delayed (HI processed) acoustic path Relative delay and perceptual relevant contribution of each path? Agnew & Thornton 2000 P. Derleth, May 2017 Page 8
Delay in HI Comb-Filter Strength of comb-filter effect determined by: - delay (ripple density ~ 1 / Δt) - ratio of direct and delayed signal (0dB: max. ripple depth, -inf... +6dB) 1000 900 800 700 600 500 400 300 200 100 0 ripple density in Hz 0 5 10 15 20 delay in ms 5 0-5 -10-15 -20 peak & notch deviation in db peak notch 0 5 10 20 25 IG in db P. Derleth, May 2017 Page 9
Delay in HI Comb-Filter Bramslow (2010) direct and delayed path, comb-filter effect for external source P. Derleth, May 2017 Page 10
Delay in HI Coupling & Hearing Loss The comb-filter-effect is determined by: - IG range (-25..+25dB) and variablity (linear gain vs. WDRC) - Ripple density (i.e. delay) and frequency region of the ripples The perceptual relevant comb-filter-effect is in addition determined by - AC&BC part of the hearing loss (i.e. audibility of the comb-filter) - Residual frequency resolution (comb-filter resolved/unresolved) For openly fitted HI with delay in the range 5-10ms and IG in the range of 0-20dB => frequency region ~800-2000Hz P. Derleth, May 2017 Page 11
Delay in HI Own Voice For own voice signals: - additional bone conduction path with approx. -0.7ms delay and frequency and time dependent (jaw position) contribution to comb-filter - It is to be expected that acclimatization-effects come into play for altered own voice signals P. Derleth, May 2017 Page 12
Delay in HI Perceptual Effects of Delay Best perceivable in no/low reverberation conditions by NH listeners 1-15ms : coloration of sound (hollow sound) >20ms : echo-like perception Most sensitive signal is own voice and wide-band noise-like signals Less perceivable in reverberant conditions and with increasing hearing loss P. Derleth, May 2017 Page 13
Studies P. Derleth, May 2017 Page 14
Stone & Moore 1999 Subjective acceptance, 7-point scale 20 NH with 4 simulated HL & appropriate HI signal-processing (4 channel WDRC) Occluded fitting External sound (voice) 65dB Delay: 6, 13, 20, 30, 40 ms, constant across frequency IG 0-27dB Lab experiment => for the lower delays all subjective ratings <3 i.e. just acceptable P. Derleth, May 2017 Page 15
Agnew & Thornton 2000 No delay / delay direct comparison (2 values, noticeable and tolerable) Occluded fitting + slit vent Own voice Flat IG, but unknown/individually selected value 18NH expert listeners Constant delay across frequency was investigated Lab experiment => detectable delay 3-5ms => acceptable delay below 10ms P. Derleth, May 2017 Page 16
Stone & Moore 2002 Subjectve rating, word production rates, speech level, f0 and f0 variation 32 SH (various losses) Occluded fitting IG ~0dB Linear processing & 3 channel WDRC Own voice Dry and live acoustics Delay: 7, 18, 30, 43ms constant across frequency Lab experiment => delay s lower than 20-30ms seem acceptable P. Derleth, May 2017 Page 17
Stone & Moore 2003 Delay only <2k (see figure) untypical for open fitting as of today 800-2000Hz transition region for delay 10 SH (moderate loss), IG according NAL, i.e. 10-20dB effective IG range 800-5800Hz Subjective rating, VCV identification speech production => delay variation across 0.8-2kHz of 9-15ms led to poor speech identification and subjective disturbance P. Derleth, May 2017 Page 18
Groth & Birkmose 2004 10NH & 10SH Non-occluding IG 10dB constant & flat (audibility was given for all signals and losses) direct sound <800Hz, delay: 2, 4, 10 ms, 800Hz ~7kHz Own voice, external speech and music Lab experiment => delays up to 10ms acceptable for NH and SH => shorter delays are prefereable (at least for NH) P. Derleth, May 2017 Page 19
Stone & Moore 2005 Subjective rating (7-point scale) & speech production rates 4 channel WDRC Delay 13, 21, 30, 40 ms, constant across frequency 25 SH with different losses IG determined by CAMEQ => -2.. 20 db (loss dependent) RT60 50ms Occluded fitting Lab experiment => just acceptable ratings (3 and better) for the shorter delays P. Derleth, May 2017 Page 20
Stone & Moore 2008 Subjective disturbance scale 1-7 NH (Simulated SH) External source open fitting Linear gain Delay constant across freqency (2, 4, 7, 12ms) IG constant => the combination of IG and delay influences disturbance level P. Derleth, May 2017 Page 21
Stone & Moore 2008 Delay constant across freqency (2, 4, 7, 12ms) IG ~0dB in the transition region around 1k or 2k and constant well above transition region => the variation of IG and delay influences disturbance level P. Derleth, May 2017 Page 22
Bramslow 2010 Subjevtive preference in A/B comparison of sound quality 18NH & 18SH (mild losses) Open fitting (& delayed path high-pass filtered 500, 1k, 2.2kHz) External sounds (speech, music, natural noise-like sounds) and own voice IG 0dB..15dB Delay: 2, 5, 10ms => delays of up to 10ms acceptable for SH listeners with mild loss => increasing high-pass filtering of delayed sound prefered for NH P. Derleth, May 2017 Page 23
Bramslow 2010 P. Derleth, May 2017 Page 24
Summary & Discussion Only a few studies used openly fitted devices with direct sound dominating in the low frequency region The range of delays tested for the openly fitted conditions is small The variation of IG (WDRC) was not directly monitored i.e. the amount/variation of comb-filter is not known in detail Some experiments were done with simulated SH, i.e. really a good representation of a hearing impairment? All lab experiments, i.e. no field trial with the chance for acclimatization => the only long-term filed trial we do is the products in the market P. Derleth, May 2017 Page 25