Psychoacoustics Study Guide PDF

Table of Contents Classical Psychoacoustic Methods 3 Signal Detection Theory 5 Psychoacoustics Crash Course 7 Masking 9 Critical Bands 11 Intensity Coding 13 Pitch Perception 17 Tuning Curves 20 Temporal Coding 21 Binaural Hearing 23 Psychoacoustics of Hearing Loss 27 Classica...

Table of Contents Classical Psychoacoustic Methods 3 Signal Detection Theory 5 Psychoacoustics Crash Course 7 Masking 9 Critical Bands 11 Intensity Coding 13 Pitch Perception 17 Tuning Curves 20 Temporal Coding 21 Binaural Hearing 23 Psychoacoustics of Hearing Loss 27 Classical Psychoacoustic Methods Method of Limits - Intensity is adjusted up and down until the patient responds - Threshold: the lowest intensity responded to 50% of the time - What affects responses: - Instruction (encouraged to guess: increases false positives, told not to guess: lowers false positives but increases misses and threshold) - Anticipating stimulus - Motivation/attention - Environmental noise, body noise, and neural noise (variable spontaneous firing rate that needs overcame for audibility) - Because of false positives and negatives and other effects, several trials are given - Response latency is used to judge if true/false positive---if it is uncharacteristically short or long, flag as false - Descending runs tend to have lower thresholds - You can average ascending and descending runs like in Hughson-Westlake (this also helps with malingering)  Method of Adjustment - Patient adjusts intensity, smoothly, themselves - Bekesy audiometry: hold the button until you can/can't hear the tone---threshold is the average of reversals. Stimulus can be sweep-frequency or fixed-frequency - Similar results/responses to method of limits Method of Constant Stimuli - A constant number of stimuli per intensity with be presented and the response (or failure to respond) will be recorded at each. Results are tallied and plotted so 50% can be found. - More trials= less effect from influences Forced Choice - Two-alternative forced choice: yes or no - Does not control for guessing - Two-interval Two-alternative forced choice: the sound will be played at A or B, tell which - Can offer more intervals (three is common) - Forcing a guess eliminates bias that comes from an individual's willingness to guess - Threshold is between certainty and chance, so 75% for 2fc and 67% for 3fc Adaptive Procedures - Response determines the next stimuli level - Method of limits is adaptive, Method of constant stimuli isn't - Start with large step-sizes and bracket threshold using increasingly small stimulus level adjustments Scaling Procedures - Magnitude estimation - Assign a number to a stimulus - May anchor response by giving a reference stimulus - Ex. Loudness scaling for MCL and LDL - Magnitude production - Subject is given control over the stimulus - Ex. Adjust loudness until it is a 50 - Fractionation - Subject makes the stimulus some fraction of the original - Ex. Halve the loudness - Cross-modality matching - Using a visual image to help a subject make a judgement about a stimulus - Ex. Adjust the line length to represent loudness Signal Detection Theory Basics - Experiments that record, for each signal level, hits, misses, false positives, and correct rejections as tallies---from this, one can derive a mathematical calculation of the magnitude of how much greater the subject's perception of the signal is than the noise - In some SDT experiments, subject's willingness to guess is manipulated through rewards and punishments (nudging the beta) - Magnitude of a sensory event has a bell-shaped (Gaussian) normal distribution - Area under the curve is 100% - X-axis is in standard deviation units (aka z-scores) - Hit + miss= 100% - Correct rejection + false positive = 100% Key Points: - When listening, there is always some background noise, so the task is always to detect signal in noise - Perception of sound is the magnitude of the sensory event - Magnitude of sensory event isn't always constant - Outside factors like neural noise and environment noise vary Terminology: - Sensitivity: how often a test diagnoses a health problem (hits) - Specificity: how often a test is correct when it has normal findings (correct rejections) - D-prime: is used to understand the strength of perception and understand the accuracy of diagnostic test results - Distance between means of distributions (SN and N) in z-scores - Ability to tell if a signal is present - D^1^ = 0: system is not sensitive to signal presence, just chance performance - D^1^ \> 5: large, greater than 5 standard deviations apart, very sensitive to signal presence - B: response bias - Independent of d^1^ - Response proclivity - Depends on: instructions, probability of stimulus, rewards/punishment, attention, and motivation - B\< 1: liberal bias - B= 1: not biased (ideal listener, decision criteria at the intersection of SN and N) - B\>1: conservative bias - PC: performance at selected bias - PCmax: performance if unbiased Receiver Operating Curve (ROC) - Keep the experiment the same but alter the subject's decision-making point (DC) - A plot of the same subject's this versus false alarm rates - A subject's responses should fall along the same d^1^ scale while altering their DC in following trials  Psychoacoustics Crash Coarse Masking - Critical band is key for masking - Masking is linear - Temporal masking: backward (masker follows the probe) and forward (probe follows masker) - Psychophysical tuning curves: steady probe with varied masker level and frequency---similar to neural tuning curves, and reflects asymmetry of the traveling wave - Upward spread of masking affects the audibility of higher frequencies Critical Bands - Minimum bandwidth of masking noise to mask probe tone - Width is the equivalent rectangular bandwidth - Normal distribution: most important frequencies are those closest to the probe frequency - Found via notched-noise and band-widening experiments - Higher frequencies have larger critical bandwidth than low frequencies - About 1/3 octave wide - Wider CB= less ringing - Auditory filter bank: series of band-pass filters - About 24 critical bands - Frequency specificity created by filters helps with sound processing and loudness perception - Waves in multiple CBs= greater loudness Binaural Hearing - Binaural summation (3 dB boost), squelch, and redundancy - Localization and lateralization - ITD: low frequency - IID: high frequency - Binaural masking level differences: altering signal and noise phase/correlation leads to variable release from masking Intensity Coding/Loudness - We hear best in the mid frequencies because of the head transfer function, OE/EAM resonance and shaping, and ME mass/stiffness - Higher threshold required for low frequencies - Phon scale: is equal loudness scale based on 1000 Hz (ex. 20 phons is loudness equal to 1000 Hz at 20 dB SPL) - Faster loudness growth (paired with higher thresholds) at low frequencies - Sone scale: loudness ratio based on 1000 Hz 40 dB SPL tone being 1 sone - Loudness growth - 10 dB SPL = doubles loudness (near miss to Weber's Law) - Below 40 dB, 6 dB SPL = doubles loudness (due to cochlear amplifier, Steven's law) - JND - For \ - We can hear 20-20,000 Hz, tonality bounds are 20-8,000 Hz - Higher frequency= more cycles needed for tonality - 250 msec needed for full tonality - Below 4000 Hz phase-locking is key, above 4000 Hz place-coding is key - Larger JND needed as frequency increases (below 4000 Hz it is a near miss to Weber's law, then JND needed rapidly increases) - Higher intensity: lower frequencies sound lower, higher frequencies sound higher Temporal Coding - We can detect a 0.001 msec change in signal - Plays a role in frequency differentiation - Temporal integration is how a system takes advantage of longer duration stimuli - Threshold worsens with duration shorter than 200/250 msec - Gap detection threshold: shortest pause in the signal that is detected as present - Frequency dependent: Humans are better at detecting gaps in high frequency stimulus than low (due to CB wideness and ringing) - Poorer close to threshold, improves up to 60 dB - We can detect 1- to 2- msec difference in starting times of two tones, but we need 20 msec difference to be able to identify which came first Hearing Loss - Articulation Index gives measure of speech understanding based off audiometric results (limited in reality) - More rapid loudness growth (recruitment) - Reduced temporal integration= inability to hear low-level, brief stimulus - Greater intensity sensitivity closer to threshold (smaller JND close to threshold compared to NH) - Widened critical bands - Due to OHC loss (tuning/sharpening) and IHC loss (impaired resolution) and nerve/synapse damage (less precision) - Greater susceptibility to masking/noise - Less sharp frequency response, pitch perception, and timing cues - Worse gap detection (reliant on ringing, low frequency regions and confused by excessive intensity fluctuation sensitivity from recruitment) - Cochlear dead regions lead to off-frequency listening (can be assessed with TEN test) - Dead region seen as 35 dB/octave slope on audiogram Masking Tone-on-tone Masking - Probe and masker are close in frequency= good masking - Probe and masker are far in frequency= bad masking (especially if the masker is higher in frequency than the probe) - Asymmetric shape of the traveling wave and the fact that high frequency maskers don't travel far up the cochlea explain why high frequencies can't effectively mask lower probes - Note: of probes and maskers are too close then beats are created - Loud sounds create harmonic distortion, and their harmonics create their own masking - Masking is linear. Turn up the masker 10 dB is 10 dB more of masking Upward Spread of Masking - High masking levels lead to greater masking of high frequency tones Temporal Masking - Forward masking: masker comes before the probe signal - Noise preceding the signal elevates the threshold of the probe signal (masks it) dependent on the length of the silent interval between the masker and the probe - How? - If the basilar membrane is still ringing from the masker, it is harder to encode a low-intensity probe tone - Also, the efferent MOC suppression may linger following the higher intensity masker - Forward masking results in a sharper psychophysical tuning curve since it reduces the ability of the listener to process information with frequencies adjacent to the probe frequency, so it better matches neural tuning curves and is used for psychoacoustics experiments - Backward masking: masker follows the probe signal - Presence of the masker a few milliseconds after the probe (silent interval \