Thread by @Dolm_in_NRW

@translationtalk Ok, so here are some excerpts from my notes that are hopefully intelligible and complete enough to be informative. I'll post them one by one in case anyone wishes to weigh in.

@translationtalk 1)
Hearing & understanding half of the time means #terps can keep up, but it’s very difficult with high cognitive load > however, gisting possible

75% intelligibility: signal somewhat better, cognitive load reduced, but listening environment still difficult > #terps struggling

@translationtalk 2) Ability 2 hear & understand exists @ around 75 % although other ppl may be around or background noise is present. Task highly influenced by signal to noise ratio (how much background noise there is). A mere 10 db difference separates “no can do” from “perfect intelligibility”!

@translationtalk 3) Absence of visual cues (due to mask or no video, say) deducts approx 6 dB (shifts psychometric function). #terps need much better signal-to-noise ratio to understand.

@translationtalk 4) Acoustic signal hitting just the one ear rather than both = loss of binaural summation. Ideal hearing task = signal hitting both ears equally (equates to 3-6 dB better sound). See above: 10 dB can make or break your task!

@translationtalk 5) Inner ear is like piano with 20k keys (1 key = 1 hZ). Full set of keys: full frequency range perceptible. Hearing cells can die off due to stress (caused e.g. by too-tasking conversion of acoustic waves 2 information). By-products of this process are toxic, causing cell death.

@translationtalk 6) Hearing loss: loss of sensitivity in the mid to higher pitch ranges. Consonants are harder to hear. Without consonants, clarity of speech is lost. Nasals/gutturals are still audible. Sibilants and plosives are lost.

@translationtalk 7a) Especially in female voices, higher frequencies (above 7k Hz) carry lots of auditory information that may be lost if not audible.

@translationtalk 7b) Super wideband above 14k hz enables hearing under more degraded circumstances but is very difficult to produce and transmit online. It requires a huge amount of bandwidth and super fast transmission speeds.

@translationtalk 8a) Not just frequency range that determines intelligibility
but also bitrate and resolution (meaning, how faithfully does a computer reproduce the original waveform).

@translationtalk 8b) -- sampling rate > pitch distortion & loss of higher frequency sounds > worsens intelligibility. ++ sampling rate > better representation, less (not no!) distortion.

Sound compression cuts off peaks & troughs > digital distortion. Headroom from lowest 2 highest inadequate

@translationtalk 9) Cognitive load:
- binaural summation lost if only one earpiece used > always use both ears!
- ambient sound (SNR loss)
- full bandwidth > 20k Hz vs wideband (approx. 50-7k Hz) vs narrowband
- reduced bitrate (poor temporal cues), slow connection, poor transmission engineering

@translationtalk 10) Non-auditory consequences of exposure:
- headache
- tinnitus
- fatigue
- poorer task performance
> potential for increased signal level (= turn up volume if you only have one control!) but this can increase noise/worsen SNR as well, boosting both signal AND distortion.

@translationtalk 11) Conclusions

- Wider band and highest bit rate (high-def audio, both ears covered
- Closed-back earphone (blocking ambient sound)
- Monitor at lowest comfortable level
- Annual audiograms (!)

@translationtalk 12) It is imperative for #terps to hear own voice alongside speaker’s for monitoring purposes. So often they sacrifice binaural summation (only one ear) and may increase volume in a move to improve signal-to-noise ratio. Consequences: see above.

@translationtalk 13) Arguable that seeing speakers' faces as if they were F2F ahead (lips, expression etc.) and widest band at least up to 14 k hz and 32 bit rate or higher = almost like IRL - then monitor can go down to lowest comfortable level (extremely important!).

@translationtalk @threadreaderapp unroll