INTRODUCTION
The language consists of public events (observable and measurable physical stimulus, composed of sounds or strokes) and private events (inferred, are representations mental meaningfully). The listener / reader has to transform physical stimuli speaking / writing that receives a representation of sounds / letters that make up the verbal issuance.
differences between the perception of oral and written language:
| Spoken Language Written Language | |
| - The sound decays with time | - No time constraints |
| - Speech is a continuous stimulus | - The script is divided into units, words, etc. |
| - affects | hearing - affects vision |
Properties the human system of perception of language consistency, flexibility and automaticity
Speech perception remains constant even when certain physical properties vary stimulation (recognizable as the same sequences with different frequencies speaking, accent, etc.) showing perception system is highly flexible and adaptive, while automatic.
FUNCAMENTALES PROBLEMS IN THIS COUNTRY (PERCEPTION OF SPEECH)
perceptual constancy in the task of acoustic-phoneme conversion:
Speech perception is a process that allows the transformation of a pattern of acoustic energy into a mental representation of the configuration stimulate (phonemes and sounds) that produce that energy.
The perceptual system extracts of these changes sound a perceptive records that correspond to linguistic units (phonemes).
Physical properties of speech sounds. Spectroscopy as a technique of analysis:
The spectroscopy allows us to examine the physical properties of sounds. Lets get a visual representation of speech or spectrogram shows the frequency composition voice in time units.
Some sequences of speech sounds are decomposed into frequency bands (expressed in hertz) called formants. Constant formant formant transition (which is curved, in which the frequency value is progressively modified) and stable part of the formant (which is flat).
This technique has shown that each phoneme is not always for the same set of acoustic units, giving rise to the problem of segmentation and lack of invariance.
The spectroscopy allows us to examine the physical properties of sounds. Lets get a visual representation of speech or spectrogram shows the frequency composition voice in time units.
Some sequences of speech sounds are decomposed into frequency bands (expressed in hertz) called formants. Constant formant formant transition (which is curved, in which the frequency value is progressively modified) and stable part of the formant (which is flat).
This technique has shown that each phoneme is not always for the same set of acoustic units, giving rise to the problem of segmentation and lack of invariance.
segmentation problem and the lack of invariance:
Delivering a sequence of phonemes, each phoneme is not articulated separately, but the articulation organs adjust their position to produce the phoneme before and after. For example, when pronouncing the phoneme / n / the position of the tongue is more retracted in co [n] contrast, more in touch with the alveoli in co [n] ato, and more at rest in co [n] call notice. The concept of demand
co-articulation of speech is the source of the problems of segmentation in the absence of invariance, of the inseparable presence of speech as fundamental frequency noise characteristics that sets the tone for you and the intensity of speech, and speech perception remains intact even when there is a loss of acoustic and noise effect.
- Segmentation: speech signal is continuous, while sounds speech is perceived as discontinuous. For example, if a separate syllable the consonant portion of the spectrum of vocal spectrum and present the subject only the portion of the consonant, it is able to guess the vowel. Lack of invariance
- : lack of correspondence between fragments of acoustic signal and discrete phonemes. Speech segments are continuous, influenced by the acoustic context. No invariant properties, listeners nevertheless capture the perceptual records and identify the sounds. That is, no physical signal characteristics identical to the same phonemes as the context signals are different, but the listener identifies the same phoneme. Concept
Delivering a sequence of phonemes, each phoneme is not articulated separately, but the articulation organs adjust their position to produce the phoneme before and after. For example, when pronouncing the phoneme / n / the position of the tongue is more retracted in co [n] contrast, more in touch with the alveoli in co [n] ato, and more at rest in co [n] call notice. The concept of demand
co-articulation of speech is the source of the problems of segmentation in the absence of invariance, of the inseparable presence of speech as fundamental frequency noise characteristics that sets the tone for you and the intensity of speech, and speech perception remains intact even when there is a loss of acoustic and noise effect.
BASIC PROCESSES IN THE PERCEPTION OF SPEECH
stages, based on linguistic considerations and not psychological. This poses problems:
peripheral auditory analysis: occurs decoding speech signals in the peripheral auditory system. Decoding mechanisms are of two kinds:
- Neuroacústicos: for example, firing patterns of nerve fibers that tune with attributes of the speech signal have been present as the first consonants of the syllable / pa /, / ga /, etc..
- Psiocacústicos: more abstract and independent of its physiological correlates. For example, band-pass filters, which perform signal processing by analyzing its components.
acoustic-phonetic analysis: We performed a linguistic processing of the signal. It involves identifying the speech segments or phonemes. The acoustic cues are coupled to the phonetic features that are abstract representations mediating between the physical planes (acoustic) and linguistic (phonetic).
- can discover the evidence of perception (perceptual categorization of speech), allowing us to identify discrete sounds, resolving the problems of segmentation and variability.
- exist at this level feature detectors or a specialized neural mechanisms in the identification of phonemic distinctive features (voicing, nasality, etc..).
- At this level certain phonetic distinctions become allophonic variations of the same phoneme, explaining certain phenomena of assimilation and phonetic processing.
- As a result of phonological analysis shows a linear sequence of phonemes, whose components are organized hierarchically: the beginning or onset (initial consonant cluster optional), rhyme or rime and coda (ending consonant optional).
No agreement on the psychological reality of these stages.
No agreement on the time course of the four processes and their interactions.
theoretical approaches derived from the consideration of the phonetic and phonological representation as linguistic constructs:
Some authors deny the existence of an independent level of phonetic representation, because the difficulty of finding a link between acoustic cues and phonetic segments. Choose to defer the settlement of the lack of invariance to a higher level of processing (lexical access process). This can lead to the conclusion that if we eliminate the phonetic level there will be no reason to postulate a phonological processing, as it would have no representation at that level input. It is considered that the phonetic and phonological representations play a role not psychologically necessary for processing the signal, but a posteriori, when the listener uses this information.
Others focus on the explanations that postulate the transformation itself acoustic-phonetic. These representations are not present in the speech signal, but are supplied by the recipient to information from his memory. Thus, although the processes of acoustic-phonetic processing are initially targeting signal properties (bottom-up process), also depend on the use of superior information (top-down process).
Some authors deny the existence of an independent level of phonetic representation, because the difficulty of finding a link between acoustic cues and phonetic segments. Choose to defer the settlement of the lack of invariance to a higher level of processing (lexical access process). This can lead to the conclusion that if we eliminate the phonetic level there will be no reason to postulate a phonological processing, as it would have no representation at that level input. It is considered that the phonetic and phonological representations play a role not psychologically necessary for processing the signal, but a posteriori, when the listener uses this information.
Others focus on the explanations that postulate the transformation itself acoustic-phonetic. These representations are not present in the speech signal, but are supplied by the recipient to information from his memory. Thus, although the processes of acoustic-phonetic processing are initially targeting signal properties (bottom-up process), also depend on the use of superior information (top-down process).
THEORIES OF ACOUSTIC-PHONETIC INTEGRATION
motor theory of speech perception (Lieberman et al.)
phonetic identification (perception signal) is performed by a specialized processing system in the perception of speech sounds than the system used in the perception of other auditory stimuli, which determines a particular processing mode (mode of speech) that tunes the auditory signals the acoustic properties and is matched to a code where the phonetic structure of language is imposed on the acoustic properties. This code is defined in terms of the properties coarticulatorias articulatory and sound. That is, there is a direct link between the systems of perception and speech production that allows the listener to determine what makes articulatory gestures the speaker and thus what produces phonetic segments. Therefore, the central idea is that speech is perceived by our tacit or unconscious knowledge of how that occurs. Description
analysis by synthesis mechanisms:
The basic mechanism of perception in this theory is the analysis by synthesis, in which the analysis refers to the process of extracting information from the signal, and synthesis processes internal signal generation from advanced acoustic cues and knowledge of the articulatory properties of sounds. This mechanism explains the problems:
According to this theory of speech perception requires no specialized processing system, but that speech is perceived by the same mechanisms as any other auditory stimulus. This is not a unified paradigm, but brings together a variety of models and explanations to some extent divergent.
implications of this theory:
phonetic identification (perception signal) is performed by a specialized processing system in the perception of speech sounds than the system used in the perception of other auditory stimuli, which determines a particular processing mode (mode of speech) that tunes the auditory signals the acoustic properties and is matched to a code where the phonetic structure of language is imposed on the acoustic properties. This code is defined in terms of the properties coarticulatorias articulatory and sound. That is, there is a direct link between the systems of perception and speech production that allows the listener to determine what makes articulatory gestures the speaker and thus what produces phonetic segments. Therefore, the central idea is that speech is perceived by our tacit or unconscious knowledge of how that occurs. Description
analysis by synthesis mechanisms:
The basic mechanism of perception in this theory is the analysis by synthesis, in which the analysis refers to the process of extracting information from the signal, and synthesis processes internal signal generation from advanced acoustic cues and knowledge of the articulatory properties of sounds. This mechanism explains the problems:
- Variability: is solved by means of interactive acoustic cues integration with discrete phonemic articulatory representations. Processing
- overall properties of speech (higher levels of representation), and suprasegmental structure (accents, intonation, etc..) And metric (syllabification) that influence the processes of acoustic-phonetic integration.
- Broadbent Ladatoped and the perceptual system adjusts to the acoustic characteristics of the emission source. Thus, judgments vary according to background information on the characteristics voice.
- As the phenomenon of duplex perception of formant transitions, or changes in the frequency bands of sound, are used to discriminate between different phonetic categories. Thus, when the formant transition is seen as the way speech is used to identify phonemes, if not, is the subject of an acoustic analysis in the auditory system. Studio
- babies: the listeners, including infants, use any information on the articulatory properties of speech. Visual information (image of a person speaking) interact with the hearing (acoustic signal the speaker). Effect
- of McGurk: when a perceiving subject is presented with conflicting auditory and visual stimuli phonetically is unconsciously adopt a compromise between the two sources of stimulation. Theory
According to this theory of speech perception requires no specialized processing system, but that speech is perceived by the same mechanisms as any other auditory stimulus. This is not a unified paradigm, but brings together a variety of models and explanations to some extent divergent.
| MOTOR COMPARISON BETWEEN THEORY AND THEORY HEARING | |
| MOTOR THEORY THEORY | |
| HEARING - The phonetic identification system is by way of speaking, it is specific to particular stimuli and different from other stimuli. | - The phonetic identification is carried out by general auditory mechanisms. It is not a unified paradigm, so each model can propose different mechanisms. |
| - The perception of speech is domain-specific and species-specific. | - rejects the idea of \u200b\u200bspeech perception and specific domain and species. |
| - Apply an analysis by synthesis mechanism for the removal of the signal and internally generated sounds. | - Reject the mechanics of analysis by synthesis by more analytical. Earlier processing of the signal in hearing levels. |
implications of this theory:
- There are authors for whom the speech signal is not variable, but there are invariant properties that allow a link between physical stimuli and microstructural phonemic representations.
- Others argue that these invariant properties arise in macro-level, where the speech signal stimulates speech patterns neurosensory representing lexical forms of memory. Model
- lexical representations consist of spectral templates (the representation of an ideal sequence of acoustic cues). Half the listener receives the speech chain fragments, is computing on the go signal spectral representations (diphones). The key feature (both diphone templates as) is that it is influenced by the context. This solves the problem of lack of invariance.
- representations do not correspond to discrete phonetic units. Thus, there is no level of representation phonetics / phonology, but the noise level has direct access to the lexical level: lexical access from spectra.
- phonemic segmentation. No
- invariance. Restriction
- suprasegmental lexical knowledge.
- syntactic and semantic variables.
Due to restrictions of the coarticulation, speech is perceived continuously. It is an active process information determined by physical, linguistic and extralinguistic. This has certain consequences:
The acoustic signal processing need not be exhaustive (identifying each and every one of the lexical segments of sensory input) to access and recognition levels of words.
From phonetic levels is an interaction between the processes of phonetic segments identification from acoustic cues and lexical access from phonological representations.
Example of interaction between perceptual processes and linguistic processes
higher order:
knowledge of prosodic and phonological peculiarities of Castilian phonetic sequences, such as content words have a stressed syllable / allows the listener to anticipate sharp acoustic events.
information is likely to higher order (lexical, syntactic, semantic) also involved in the primary processes of signal analysis. In the study presented Picket Pllack and subjects of conversation fragments (words from a sentence) and asked to identify those words. Not exceed 50% of correct words. Appears because the speech signal remains unintelligible until the recipient does not have enough background information to formulate hypotheses on the pair phonological content of the message.
knowledge of prosodic and phonological peculiarities of Castilian phonetic sequences, such as content words have a stressed syllable / allows the listener to anticipate sharp acoustic events.
information is likely to higher order (lexical, syntactic, semantic) also involved in the primary processes of signal analysis. In the study presented Picket Pllack and subjects of conversation fragments (words from a sentence) and asked to identify those words. Not exceed 50% of correct words. Appears because the speech signal remains unintelligible until the recipient does not have enough background information to formulate hypotheses on the pair phonological content of the message.
Three tests show the influence of the higher processes of recognition of basic mechanisms of perceptual processes:
The speech recognition processes are open to influences from higher levels of processing (suprasegmental information-metric-prosodic, lexical, syntactic and semantic) that imposes restrictions continuous talk, and operate automatically and unconsciously.
It can not be concluded that speech perception is a process distinct from the perception of isolated sounds, but a process determined by the physical constraints of the signal and constraints of linguistic representations that are recovered at higher levels.
- phoneme restoration effect: phonetic substitution unconscious material, absent from the acoustic signal, a stimulus not present in the speech signal. Restoration
- errors (tracking task), the subject unconsciously replaced erroneous speech stimuli for correct forms. It seems that this happens with both lexical and syntactic information and phonetics. Phenomenon
- selective listening: when a subject is subjected to a dichotic listening task are recorded effects caused by the material submitted by the unattended ear (the subject does not hear the meaning of the message, the message is not addressed may cause interference with the task, etc.).
The speech recognition processes are open to influences from higher levels of processing (suprasegmental information-metric-prosodic, lexical, syntactic and semantic) that imposes restrictions continuous talk, and operate automatically and unconsciously.
It can not be concluded that speech perception is a process distinct from the perception of isolated sounds, but a process determined by the physical constraints of the signal and constraints of linguistic representations that are recovered at higher levels.
0 comments:
Post a Comment