Тексты на английском
<<  Documentation Utilizing WVEIS On the Web (WOW) December 2, 2011 The suez crisis 1956  >>
Sound Localization in Natural Hearing
Sound Localization in Natural Hearing
Example of Clarity for Speech
Example of Clarity for Speech
Example of Clarity for Speech
Example of Clarity for Speech
What happens in a room
What happens in a room
Sound separation: Localizing a String Quartet
Sound separation: Localizing a String Quartet
ILD differences in human hearing
ILD differences in human hearing
Stereo Reproduction
Stereo Reproduction
With broadband signals the perceived position is closer to the left
With broadband signals the perceived position is closer to the left
Sound diffracts around the head and interferes with sound from the
Sound diffracts around the head and interferes with sound from the
Sound panned to the middle between left and center is perceived beyond
Sound panned to the middle between left and center is perceived beyond
Delay Panning
Delay Panning
Compare Soundfield and Binaural
Compare Soundfield and Binaural
Картинки из презентации «Pitch, Timbre, Source Separation, and the Myths of Sound Localization» к уроку английского языка на тему «Тексты на английском»

Автор: Sony Customer. Чтобы познакомиться с картинкой полного размера, нажмите на её эскиз. Чтобы можно было использовать все картинки для урока английского языка, скачайте бесплатно презентацию «Pitch, Timbre, Source Separation, and the Myths of Sound Localization.ppt» со всеми картинками в zip-архиве размером 3004 КБ.

Pitch, Timbre, Source Separation, and the Myths of Sound Localization

содержание презентации «Pitch, Timbre, Source Separation, and the Myths of Sound Localization.ppt»
Сл Текст Сл Текст
1Pitch, Timbre, Source Separation, and 14vary. In the author’s experience a
the Myths of Sound Localization. David binaural recording from a great concert
Griesinger David Griesinger Acoustics seat can have sharper localization over
dgriesinger@verizon.net the entire width of the orchestra than the
www.davidgriesinger.com. best recording reproduced over stereo
2Sound Localization in Natural Hearing. loudspeakers. But in a poor seat
The sensitivity of human hearing is localization is considerably worse.
weighted to the frequencies of vocal 15Stereo Reproduction. Stereo recordings
formants. These frequencies carry most of manipulate the position of image sources
the information in speech And they also between two loudspeakers by varying time
carry most of our ability to localize or level. With the common sine/cosine pan
sounds Vertical localization is almost law a level difference of 7.7dB to the
entirely above 1000Hz So elevated speakers left should produce a source image
for “3d” audio need not be large and half-way between the center and the left
heavy! Transfer function from sound loudspeaker, or 15 degrees to the left. If
outside the head through the outer and we listen with headphones, the image is at
middle ear. Notice that the pesky low least 30 degrees to the left. And the
frequencies are largely filtered away, and position is independent of frequency.
there is almost 8dB of boost at 3kHz. 16With broadband signals the perceived
3Separation of simultaneous sounds. The position is closer to the left
author’s current work concentrates on the loudspeaker.
brain processes that enable our ears to 17Sound diffracts around the head and
separate simultaneous sounds into interferes with sound from the opposite
independent neural streams. At formant speaker. Sound from the left speaker
frequencies separation requires that diffracts around the head and appears at
sounds have a definite pitch, and multiple the right ear. The diffracted signal
harmonics above 1000Hz. The phases of the interferes with the signal from the right
harmonics must not be altered by acoustics loudspeaker – and at s~1600Hz the sound
in the first 100ms. If sounds can be pressure at the right ear can be nearly
separated by pitch they can be zero.
individually localized even in the 18Sound panned to the middle between
presence of noise or other pitched sounds. left and center is perceived beyond the
But if the signal is speech or music with loudspeaker axis at high frequencies. Note
definite pitches we can easily localize that frequencies we principally use for
and understand two simultaneous speakers sound localization in a complex field the
or musicians. Houtsma found that two perceived position of the source is very
simultaneous monotone speech signals in different from the sine/cosine pan-law.
the same location can be separately The diagram shows the perceived position
understood if the pitch difference is only of a sound source panned half-way between
half a semitone, or 3%. The author has center and left with the speakers at +-45
examples of pitch separation of one degrees.
semitone on his web-site Source 19Delay Panning. Techniques such as ORTF
separation, perceived distance, clarity of combine amplitude and delay panning But
localization, and clarity of sound are ALL delay panning is even more frequency
related to the same physics of dependent. And if a listener is not in the
information. sweet spot Delay panning does not work.
4Clarity, Distance, and Audience This graph shows the approximate position
Attention. We detect “near” versus “far” of third octave noise bands from 250Hz to
instantly on perceiving a sound “Near” 2000Hz, given a 200 microsecond delay.
sounds demand attention and sometimes This delay corresponds to a 15 degree
immediate attention. “Far” sounds can source angle to two microphones separated
usually be ignored Cinema and Drama by 25cm. The panning angle is about 15
directors demand that dialog be perceived degrees at 250Hz – but it is nearly 90
as “Near” Drama theaters are small and degrees at 2000Hz.
acoustically dry Movie theaters are dry 20Stereo localization is an illusion
and use highly directional loudspeakers based on fuzzy data. The only stable
with linear phase response at vocal locations are left, center, and right And
formant frequencies. High sonic clarity center is stable only in the sweet spot.
and low sonic distance requires that Confronted with an image between center
harmonics in the vocal formant range are and left, or center and right, the brain
reproduced with their original phase must guess the location based on an
relationships Unmodified by loudspeaker average of conflicting cues. The result
characteristics or reflections. This can be beyond the speaker axis. Our
aspect of sound reproduction is not perception of sharp images between center
commonly recognised, either in live and left or right is an illusion generated
performance or in sound playback. Tests of by our brain’s desire for certainty, and
microphone techniques where the it’s willingness to guess.
microphones are beyond the room critical 21Headphone reproduction of panned
distance are of limited value! images. Headphones reliably spread a
5Example of Clarity for Speech. This panned image +-90 degrees The frontal
impulse response has a C50 of infinity STI accuracy is about +-2.5 degrees And the
is 0.96, RASTI is 0.93, and it is flat in perceived position is independent of
frequency. In spite of high C50 and frequency – if the headphones perfectly
excellent STI, when this impulse is match the listener’s ears It is useful to
convolved with speech there is a severe adjust a 1/3rd octave equalizer with 1/3rd
loss in clarity. The sound is muddy and octave noise bands to obtain a stable
distant. The sound is unclear because this central image. Thus it is possible to hear
IR randomizes the phase of harmonics above precisely localized images from pan-pots
1000Hz!!! when listening through headphones But not
6Demonstration. The information carried with speakers. I must emphasize that
in the phases of upper harmonics can be headphone reproduction requires accurate
easily demonstrated: Dry monotone Speech equalization of the headphones – ideally
with pitch C Speech after removing individually different for the left and
frequencies below 1000Hz, and compression right ear!
for constant level. C and C# together. 22Coincident Microphones. Coincident
Spectrum of the compressed speech. It is microphones produce a signal similar to a
not difficult to separate the two voices – pan-pot. But the range of the pan is
but it may take a bit of practice! limited. The most selective pattern is
7What happens in a room? Measured figure of eights at 90 degrees (Blumlein)
binaural impulse response of a small A source 45 degrees to the left will be
concert hall, measured in row 5 with an reproduced only from the left loudspeaker.
omnidirectional source on stage. The A 3.3 degree angle from the front produces
direct level has been boosted 6dB to a 1dB level difference in the outputs. Not
emulate the directivity of a human as good as natural hearing, but not too
speaker. RT ~ 1s Looks pretty good, bad when listening through headphones. To
doesn’t it, with plenty of direct sound. produce a “full left” or “full right”
But the value of LOC is -1dB, which signal the peak sensitivity of one
foretells problems… microphone must lie on the null of the
8Sound in the hall is difficult to other microphone. Cardioid microphones can
understand and remember when there is just only do this if they are back-to-back (180
one speaker. Impossible to understand when degrees apart) To pick up a decorrelated
two speakers talk at the same time. All reverberant field the microphones must be
these effects depend on the coherence of at least supercardioid. Cardioid
upper harmonics. When sound is reproduced microphones pick up a largely monaural
over multiple loudspeakers this quality reverberant field at low frequencies in
usually suffers. This difficulty applies the standard ORTF configuration. If
both to Ambisonics and WFS, especially properly equalized hypercardioid or
because spatial aliasing is significant at supercardioid microphones sound much
formant frequencies. C in the room C# in better.
the room C and C# in the room together. 23Compare Soundfield and Binaural. When
9Sound separation: Localizing a String you use headphones to compare a binaural
Quartet. From the author’s seat in row F recording to a Soundfield recording with
behind the lady in red the string quartet any pattern: The soundfield has a narrower
was +-10 degrees in width. But in the front image – (it is not as sharp) more
presence of substantial reverberation it reflected energy – (there is no head
was possible to distinctly localize all shadowing) And less envelopment – (the
four players with eyes closed, even when direct sound is less distinct) Comparing a
they played together. This implies a binaural recording on headphones to a
localization acuity of better than three Soundfield recording on speakers is not a
degrees. With just slightly more reasonable thing to do!
reverberation it was not possible to 24Ambisonics and WFS. Ambisonics has a
localize the musicians at all. problem with head shadowing. The lateral
10Perception of reverberation and velocity vector is created by subtracting
envelopment. The goal of the ear/brain is the left loudspeaker sound from the right
to extract meaningful sound objects from a loudspeaker sound. But if the head is in
confusing acoustic field. To the brain the way the signals do not subtract – and
reverberation is a from of noise. Where the lateral sound is perceived stronger
possible the brain stem separates direct than it should be – and often as excess
sound from reverberation, forming two pressure. Even in high order Ambisonics
distinct sound streams: foreground and the frequency dependent panning problems
background. Perceiving reverberation and still exist between loudspeakers. This
envelopment is only possible when the results in a lower Direct to Reverberant
direct sound can be separately ratio at frequencies above ~500Hz. Gerzon
perceived!!! Clarity of the front image is knew all about the problem, switching to
a vital part of this process. When the crossed hypercardioid patterns above
front image is muddy reverberation becomes 500Hz. The resulting directional beams are
a part of the front image and cannot be about 100 degrees wide! And the front
localized. These facts are unappreciated image has all the problems of stereo. If
in current concert hall design. Almost more speakers are used accuracy of
invariably a recording has a clearer front localization is not increased, as the
image than a typical concert seat, where first order patterns are not selective
“well blended” sound is all you can hear. enough to limit reproduction to only two
It need not be so. With good acoustic speakers at the same time.
design a concert seat can have better 255.1 Front. A precise center location
clarity and envelopment than any recording is perceived anywhere in the listening
reproduced over loudspeakers. space, and not just at the sweet spot
11Localizing separated sounds in natural (sweet line). Accuracy of localization in
hearing. It is well known that we localize the front is greatly improved As long as
sounds through: the Interaural Level only two speakers are active a time. This
Difference (ILD) and the Interaural Time requires panning from center to left or
Difference (ITD) Experiments with sine center to right, and not simultaneously
tones show that ITD is not useful above producing a phantom image from the left or
2kHz due to frequency limits on nerve right speakers. An image panned to the
firings. And that ILD loses accuracy below middle from center to left is perceived at
1kHz as head shadowing decreases. But high 30 degrees at 3kHz. This is a factor of
harmonics in the 1kHz to 4kHz range of low two improvement over two channel stereo.
frequency fundamentals contain nearly all But the localization accuracy of three
the information of speech And also provide front speakers is still far inferior to
timbre cues that identify musical natural hearing. A five channel, five
instruments. When these harmonics are speaker front array is considerably better
present we find that we can localize tones – again as long as only two channels are
accurately with ILD To understand our active for a given sound source.
ability to localize speech and music we 265.1 side and rear. Perceiving a
need to use signals that include harmonics discrete image at the side between the
When harmonics are present our ability to front and rear speaker is difficult or
localize can be extremely acute, +-2 impossible. Sharp localization is only
degrees or better. possible if a single speaker is activated,
12ILD differences in human hearing. Note either a front or a rear. Amplitude
that the 2 to 3dB of level difference panning between the two rear speakers is
between the two ears is nearly constant in possible, with all the disadvantages of
the vocal formant range. MIT Kemar HRTF two-channel stereo.
for 0 degrees elevation and 5 degrees 27Vertical Localization. Vertical
azimuth. Head shadowing is >2dB above localization is achieved entirely through
800Hz. If we assume a 1dB threshold for timbre cues above 1000Hz. Sounds in the
level differences we should be able to horizontal plane lack frequencies in the
localize a frontal source with an range of 6000-10000Hz. Augmenting these
uncertainty of only 2 degrees. And we can… frequencies with loudspeakers above and
13Individual Instruments vs Sections. forward of the listener leads to a
Ability to localize individual instruments pleasing sense of natural reverberation.
with an uncertainty of 3 degrees or better Two loudspeakers above the listener
is possible in good acoustics. The ability reproducing decorrelated reverberation
disappears abruptly when there are too sound much better than one. The best
many early reflections, and multiple location is to the left and right 60 to 80
instruments fall together in a fuzzy ball. degrees above the listener. The signals
With eyes open the visual localization reproduced can be derived from the
dominates, and we are typically not aware reverberant component of a stereo or 5.1
that auditory localization is lost. When recording.
multiple instruments are playing the same 28Conclusions. Localization in natural
notes (a section) the uncertainty hearing takes place mostly at frequencies
increases dramatically. The section tends above 1000Hz, and has a precision of ~ 2
to be both visually and sonically wide. degrees in the horizontal plane. With care
But in a performance of a Haydn Symphony this precision can be reproduced over
by the Boston Symphony Orchestra under Ton headphones from a binaural or amplitude
Koopman, the string sections played panned recording. Commonly used
without vibrato. The visual image was wide pressure-gradient microphone techniques
– but the auditory image was of a single are not capable of capturing sound
instrument, and was sharply localized at direction with this precision. The vey
the center of the section. best they an do is about four degrees, and
14Summary of Natural Hearing. When early techniques that use cardioid microphones
reflections are not too strong we are able give at best about 8 degrees. Two channel
to localize multiple sound sources with loudspeaker reproduction suffers from
high precision – approximately two frequency dependence for frequencies above
degrees. If multiple people are talking 1000Hz, spreading apparent localization of
simultaneously we are able to choose to a panned image over ~30 degrees. The brain
listen to any of them. If multiple must make a best guess for the perceived
instruments are playing we are able to position of a source – and the frequency
follow the lines of several at the same spread is sometimes audible. Multichannel
time. These abilities disappear abruptly reproduction in 5.1 improves the accuracy
when the early reflections exceed a about a factor of two. The more channels
certain level with respect to the direct used the more natural the frontal image
sound. The information responsible for becomes, as long as only two adjacent
these abilities lies primarily in loudspeakers are active for any one
harmonics above 1000Hz from lower source. First order Ambisonics with four
frequency tones. Localization for natural speakers has all the problems of stereo
hearing is independent of frequency!!! and then some. With multiple speakers
Acuity, the sharpness of localization, can localization accuracy is slightly
vary, but the perceived position does not improved.
Pitch, Timbre, Source Separation, and the Myths of Sound Localization.ppt
cсылка на страницу

Pitch, Timbre, Source Separation, and the Myths of Sound Localization

другие презентации на тему «Pitch, Timbre, Source Separation, and the Myths of Sound Localization»

«The english-speaking countries» - Great Britain. Disneyland. Australia. The English-speaking countries. USA. Scotland.

«The green movement» - National offices Green Peace are opened in 43 countries of the world as the independent units working over achievement of the purposes of the national projects. The main objective — to achieve the decision of global environmental problems, including by attraction to them of attention of the public and the authorities.


«Женщина the woman» - Наименование молодой девушки в современном английском языке. A woman’s tongue wags like a lamb’s tail. As great a pity to see a woman cry as a goose go barefoot. 9 семантических подгрупп, характеризующих женщин по: «Un homme»- франц. « A man »- англ. Женский интеллект. « Der mann»- нем. Пути пополнения лексической группы «женщина» в английском языке.

«Meat» - Meat… A MAN who eats ALL. In soy no such amino acids that are present in the meat. Meat products - a valuable source of iron, potassium, phosphorus. Useful properties of meat and meat products are undeniable. The proteins found in meat, its composition is very close to human. WHAT GOOD MEAT? meat.

«Word 2010» - Tables. Insert comments in a document. Customize track changes. Inserting merge fields. Exploring Microsoft Office Word. Nonprinting formatting marks. Font attributes. Footnotes and endnotes. Table position and alignment. Widows and orphans. Paragraph indents. Selecting recipients. Figure captions. Line and paragraph spacing.

Тексты на английском

46 презентаций о текстах на английском

Английский язык

29 тем
900igr.net > Презентации по английскому языку > Тексты на английском > Pitch, Timbre, Source Separation, and the Myths of Sound Localization