The effect of wearing conventional and level-dependent hearing protectors on speech production in noise and quiet Ghazaleh Vaziri Christian Giguère Hilmi R. Dajani Nicolas Ellaham Annual National Hearing Conservation Conference San Antonio, Texas, 23 25 February 2017 Hearing Research Laboratory
Background Occupational noise is a risk factor for the health and safety of workers Stress, fatigue Interference with communication Reducing job performance Temporary and permanent hearing loss Cardiovascular problems
Background Engineering controls are the preferred noise mitigation method, but it is not always possible Hearing protectors (s) are often used s may affect different dimensions of auditory situational awareness: Sound detection Sound localization Recognition of important sounds Speech communication
Background Talker Environment Listener Speech production Factors: Gender Vocal effort Lombard effect Lang. proficiency Hearing protection Acoustic transmission Factors: Distance Reverberation Speech perception Factors: Hearing loss Lang. proficiency Non-acoustic cues Hearing protection NOISE
Background Lombard effect: reflex of talkers to raise vocal effort in noise CHARACTERISTICS OF LOMBARD EFFECT Vocal effort Speech level Pitch frequency (F0) Formant frequencies (F1, F2) Spectral balance Speaking rate Duration Intelligibility Shift to higher frequencies Vowels: Consonants: Words: Sentences:
Background Effect of s on speech production Increases users' perception of their own voice (Occlusion effect) Decreases their perception of surrounding noise leading to a reduction in Lombard effect Alters speech output in quiet and in noise Most studies focused on conventional s in continuous noises (Kryter 1946, Howell & Martin 1975, Hormann et al. 1984, Tufts & Frank 2003, Brungart et al. 2012, Bourserhal et al. 2016) Quiet slight increase in speech level (1-3 db) Noise decrease in speech level with increasing noise level (4-11 db)
Objectives 1. Develop a method of speech signal extraction in the external noise field. 2. Investigate the effects of noise type (continuous, fluctuating) and listener situation (wearing or not) on speech production. 3. Investigate the effects of level-dependent s on speech production. NHCA, San Antonio, Texas (February 23 25, 2017)
Measurement setup Noise simulation facility Quiet and noise 55-85 dba External noise field from 6 loudspeakers Hearing protector Conventional and level-dependent Subject talkers Normal hearing Task Utter list of 11 short sentences (HINT) Example The engine is running Manikin 1 m in front Recording equipment Sound level meter (B&K Type 2235) Free-field mic. (B&K Type 4189) at 25 cm Audio Interface (TASCAM US-366) Mic. manikin Sp subject N5 N6
Measurement setup Lombard speech extraction in an external noise field Direct Waveform Subtraction (Brungart et al., 2012) Adaptive noise cancellation (Vaziri et al., 2015) Noise Subtract + Noisy speech Enhanced speech (DWS) Two-channel Adaptive Filtering (O'Shaughnessy, 2000) Enhanced speech (ANC)
Experiment 1 24 participants with normal hearing (12 male, 12 female) Conventional passive earmuff 3M PELTOR Optime 98 with a Noise Reduction Rating (NRR) of 25 db Four different ear conditions: Talker Open Open Listener (Manikin) Open Open Speech recording in quiet and 4 different noises (Pink, IFFM, SSnoise, Helicopter) at two different levels (70 and 85 dba)
Experiment 1 Band Level (db SPL) 60 50 40 30 0.5 0-0.5 0.5 0-0.5 0.5 0-0.5 0.5 0-0.5 0 1 2 3 4 Time (s) 20 63 125 250 500 1K 2K 4K 8K One-third octave-band frequency (Hz) Pink noise IFFM noise SSnoise Helicopter Pink noise IFFM noise SSnoise Helicopter
Experiment 1 Effect of : Talker: 0.6 db (non-significant) Speech Level (d db SPL) 85 Listener: 6-7 db 80 (significant) 75 speech in quiet Listener Open 70 Open Open Talker ear condition
Experiment 1 Effect of Noise level: Talker Listener Rise (db) Open Open 6.6 Open 4.5 Speech Level (d db SPL) 95 90 85 80 Open 2.6 75 1.8 Effect of Noise type: Lower speech levels in fluctuating noises Higher speech levels in low freq. noises 95 90 85 80 75 Talker-Open Listener-Open 70 85 Talker- Listener-Open speech in noise 95 90 85 80 75 95 90 85 80 Talker-Open Listener- 70 85 Talker- Listener- 75 70 85 70 85 Noise Level (dba) Pink IFFM SSnoise Helicopter
Experiment 1 speech in noise Effect of : Talker: 2-6 db at 70 dba 5-10 db at 85 dba Listener: 3-5 db at 70 dba 1-4 db at 85 dba Speech Level (db SPL) 100 95 90 85 80 70 dba 85 dba Listener Listener Open Open 75 Open Open Open Open Talker ear condition Pink IFFM SSnoise Helicopter
Experiment 1 Pitch Frequency (F0): Similar pitch effects for both genders 440 400 360 speech in noise 70 dba 85 dba Listener Listener Open Open Pitch increases with speech levels Pitch changes with noise level and talker/listener condition follow same patterns as speech level data Pitch (Hz) 320 280 240 200 160 120 Open Open Females Males Open Open Talker ear condition Pink IFFM SSnoise Helicopter
Experiment 2 24 participants with normal hearing (12 male, 12 female) Level-dependent 3M Peltor PowerCom Plus earmuff Talker Device setting Open - -S4 -S1 -Off Level-dependent @ High gain Level-dependent @ Low gain Passive (NRR of 25 db) Speech recording in quiet and 2 different noises (Pink and Pulse Pink) at three different levels (55, 70 and 85 dba) 0.1 0-0.1 0.1 0 Pink noise Pulse Pink (16/s) -0.1
Experiment 2 Input-output function (speech noise) for Level-dependent 3M Peltor PowerCom Plus Gain (db) + 13 + 9 + 4-2 -6 -Off -30 db
Experiment 2 Effect of : 76 speech in quiet Small and nonsignificant Trend towards lower speech levels with high level-dependent gain setting (-S4) Speech Leve el (db SPL) 74 72 70 68 Open S4 S1 Talker ear condition Off
Experiment 2 More prominent effect of the higher the noise level Lower speech levels with s than Open at 70 and 85 dba Biggest drop (4-10 db) with -Off Least drop (2-6 db) with -S4 Slightly higher speech levels in continuous Pink than Pulse Pink Speech Level (db SPL) 95 90 85 80 75 Pink Pulse Pink 55 dba 55 dba speech in noise 70 dba 85 dba 70 Open Open Open S4 S1 Off S4 S1 Off Talker ear condition S4 S1 Off
Conclusions Direct waveform subtraction and adaptive cancellation are effective laboratory methods for extracting speech when s are worn. Results show: Reduced speech levels (5-10 db) the larger the noise level when the Talker is wearing passive s. Increased speech levels (3-7 db) the lower the noise level or in quiet when the Listener is wearing s. Level-dependent protection promotes higher speech output, especially at high amplification settings, but speech levels remain lower than with open ears. Slightly lower (0-3 db) speech levels in fluctuating noises. Wearing s can affect speech production through a complex interaction of factors.
The effect of wearing conventional and level-dependent hearing protectors on speech production in noise and quiet gvazi024@uottawa.ca Ghazaleh Vaziri Christian Giguère Hilmi R. Dajani Nicolas Ellaham Annual National Hearing Conservation Conference San Antonio, Texas, 23 25 February 2017 Hearing Research Laboratory