Lecture 15 Flashcards
What is timbre?
Difference between two sounds that are otherwise the same loudness (intensity) and pitch (periodicity).
How sounds differ in spectral properties, time-varying properties.
What might the difference in spectral shape be for two sounds for timbre?
Both sounds synthesized with same formant frequencies but show different qualities due to nasal formants contributing. Spectral differences affects the perceived timbre/quality of sound.
How might perceived quality change between two sounds for timbre?
The sound wave is inverted and played backwards. Will have different sound to it.
What is the auditory scene analysis problem?
In natural listening environments, speech and other sounds of interest are more often than not heard in the presence of other sounds.
Ear is receiving combination of input of all the energy sounds waves produced around you.
The problem is, how does the brain allocate parts of the auditory input to each perceived auditory event?
In auditory scene analysis, what are the two apporoaches to grouping the auditory input?
- Group of heuristic rules (grouping principles; principles of organization) to any type of sounds (speech/non-speech). Work most of the time - they have evolved, been formed by our bodies and brains with experiences we’ve had with sounds.
- Schemas (organized mental representations of events) - knowledge about sounds we have acquired.
In the auditory scene analysis grouping processes, what are the two types of perceptual organization that take place?
Simultaneous grouping - involved assigning simultaneous/overlapping frequency components to same/different streams
Sequential grouping - involves assignment of successive frequency components to same or different streams.
Temporal characteristic not as inherent as it is for auditory events - time. Often changes over time.
What are the three types of simultaneous grouping principles?
REFER to slides Nov 28 - #24.
- Frequency - can resolve frequency components, separate components that happen at the same time, on the bases of different groups. Sounds (components) that starts before other components are grouped separate.
- Frequency - sounds that don’t fit with regularity of spectrum with other components, are also heard separately, out of that mixture of components.
- Frequency and amplitude - components modulated in frequency or amplitude following a certain pattern have one component that changes sinusoidally with a different phase - brain should be segregating the sound that is different.
Simultaneous grouping - onset synchrony: what happens when frequency components start 30ms or longer before other components?
What is the result of this?
These components are perceptually separated.
Onset disparities of 30ms between signal and masker could lower discrimination thresholds for signal.
Both for recognizing timbre of the vowel, or judging pitch of sounds. If you have components that start before other components, you are behaving all the components as if you have removed the asynchronous components.
In simultaneous grouping - onset synchrony, what happens when the timing of sounds is manipulated?
First example: continuous sound that is not the same onset as the other sound, interrupted so its onset lines up with another sound. Sounds like the alternation of two sounds - difference between the two is noticeable.
Second example: sound played continuously and the second sound does not have the same synchronous set. Sounds more similar
In simultaneous grouping, what is the result of modulation - either frequency of amplitude?
Frequency components that change at the same time and in same way are grouped together (more fused), can recognize the sounds in speech easier.
Differences in modulation pattern are not strong segregation cues except when there are modulated vs. unmodulated components.
(weak principle when you compare sounds that change in different ways)
What is the perceptual segregation result of two steady-state tones simultaneously played?
When two steady-state tones are simultaneous, the perceptual segregation can be enhanced if one is frequency amplitude modulated.
In simultaneous grouping, what is harmonicity?
Differences in F0 can be exploited because many natural sounds have harmonic structures (frequency components are multiples of F0).
Generates the regular spacing of harmonics in the spectrum - even though sound is not regular harmonic sound.
Low-numbered harmonics are resolved and can be grouped because they form a regular pattern in frequency (spectral regularity).
In simultaneous grouping, harmonicity and spectral regularity, what is the effect of different F0 on speech perception?
What is the experiment associated with this?
Participants asked to recognize words in sentence pairs presented together.
Play two sounds with different fundamental frequencies, their harmonics will segregate (heard separately). If you increase the frequency separation, the accuracy of performance increases, but decreases at 12 semitones (harmonics match even though they are in different octaves)
Recognition of ‘key’ words in simultaneous pairs of sentences enhances with F0 separation up to 3 semitones.
In simultaneous grouping what is the effect of harmonicity on detection of individual harmonics, plus on the pitch of complex tones?
What is the experiment associated with this?
Participants as whether or not they heard a separate harmonic of the rest of them, but by asking them to judge pitch of the harmonic series.
Individual harmonics: harmonic heard as separate tone if mistuned by 1-3%. (mistuning causes the pitch of the whole series to change a bit by changing the frequency)
Pitch of complex tones: if a harmonic is mistuned by 8% or larger of harmonic frequency, it stops contributing to the pitch of complex sound.
By mistuning harmonics of a series, you first are integrating the harmonic into pitch or series up to about 8% - not perceived after this point. Heard as a separate sound.
In simultaneous grouping, what is spectral regularity?
What does this have to do with mistuning?
How is harmonicity involved?
The regularity of frequency patterns does not have to be specific to multiple integer relationships: frequency components that obey any spectrally regular pattern are fused. Harmonicity is an example. If you take a harmonic series and increase the frequency of each harmonic by a fixed amount, will take series that is not harmonic, but regular pattern of frequency spacing.
Mistuning - change the frequency of a component by gradual method, the component whose frequency has changed also stands out of that sound. Sound that does not fit in with spacing of the harmonics is segregated.
