This is the generic class for apparatus and corresponding
methods for constructing, analyzing, and modifying units of human language by data processing, in which there
is a significant change in the data.
This class also provides for systems or methods that process speech signals for storage, transmission,
recognition, or synthesis of speech.
This class also provides for systems or methods for bandwidth
compression or expansion of an audio signal, or for time compression
or expansion of an audio signal.
Class 704 is structured into three main divisions:
A. Linguistics.
B. Speech Signal Processing.
C. Audio Compression.
See Subclass References to the Current Class, below, for the
subclasses located within each of these three main divisions.
SECTION II - LINES WITH OTHER CLASSES AND WITHIN THIS CLASS
A. LINGUISTICS
1. This class does not include subject matter wherein significant
details of the modification or construction of documents are claimed.
(See Class ?0? in the Search Class notes below in References to
Other Classes, regarding Document Processing).
2. This class does not include subject matter directed to significant
details of teaching languages.
(See Class 434 in the Search Class notes in References to Other Classes,
below).
3. This class does not include subject matter directed to significant
details of the construction, analysis or modification of computer languages. (See Class 717 in the Search
Class notes in References to Other Classes, below).
B. IMAGE ANALYSIS
1. This class does not include subject matter wherein significant
image analysis is performed and speech signal
processing is nominally claimed (see Class 382 in the Search Class
notes in References to Other Classes, below).
2. This class includes subject matter directed to speech signal processing disclosed or
claimed in plural diverse arts such as image analysis (classified,
per se, in Class 382).
C. AUDIO SIGNAL PROCESSING
1. This class does not include subject matter wherein nominal
bandwidth or time modifications are performed for other audio processing
defined in Classes 381 or 84 (see Search Class notes below in References
to Other Classes). Examples of subject matter not included are:
Stereo, sound effects, hearing aids, input and output transducers,
and musical instruments.
2. This class includes audio signal processing wherein significant
processing is performed to modify the signal"s bandwidth
or time characteristics for compression or expansion of the signal.
D. COMMUNICATIONS
1. This class does not include subject matter wherein significant
details of a distinct communications system or telephone link is
performed and speech signal processing
is nominally claimed (see Classes 340, 370, 375, 379, 455 in the
Search Class notes below in References to Other Classes.).
2. This class includes subject matter directed to speech signal processing disclosed or
claimed in plural diverse arts such as various types of communication
systems.
E. APPLICATIONS
1. This class does not include subject matter wherein significant
details of application systems are performed and speech signal
processing is nominally claimed.
2. This class includes subject matter directed to speech signal processing disclosed or
claimed in plural diverse arts to include electrical and mechanical
systems. Examples would include systems controlled by speech recognition, systems which create
specific displays of speech data,
systems for editing speech data
and otherwise unrelated systems which incorporate speech signal processing
details such as placing a speech synthesizer into
novelty items.
SECTION III - SUBCLASS REFERENCES TO THE CURRENT CLASS
Music,
subclasses 1+ for instruments used in producing music to include
(a) electrical music instruments, (b) automatic instruments, and
(c) hand-played instruments. Automatic and hand-played instruments
are divided into four groups: stringed, wind, rigid vibrators,
and membranes. This class also includes some accessory devices
generally recognized as belonging to the art or industry.
Acoustics, various subclasses, for mechanically transmitting,
amplifying and ascertaining the direction of sound and for mechanically muffling
or filtering sound.
Communications: Electrical,
subclasses 825+ for controlling one or more devices to obtain a plurality
of results by transmission of a designated one of plural distinctive
control signals over a smaller number of communication lines or
channels.
Coded Data Generation or Conversion, various subclasses for electrical pulse and digit code
converters (e.g., systems for originating or emitting a coded set
of discrete signals or translating one code into another code wherein the
meaning of the data remains the same but the formats may differ).
Computer Graphics Processing and Selective Visual
Display Systems, various subclasses for the selective control of two or
more light generating or light controlling display elements in accordance
with a received image signal, and
subclasses 1.1 through 3.4for visual display systems with selective electrical
control including display memory organization and structure for
storing image data and manipulating image data between a display
memory and display device.
Dynamic Magnetic Information Storage or Retrieval, which is an integral part of Class 369 following
subclass 18 , for record carriers and systems wherein information
is stored and retrieved by interaction with a medium and there is
relative motion between a medium and a transducer, for example,
magnetic disk drive devices, and control thereof, per se.
Static Information Storage and Retrieval, various subclasses for addressable static singular storage
elements or plural singular storage elements of the same type (i.e.,
the internal elements of memory, per se).
Dynamic Information Storage or Retrieval, various subclasses for record carriers and systems
wherein information is stored and retrieved by interaction with
a medium and there is relative motion between a medium and a transducer.
Multiplex Communications, for the simultaneous transmission of two or more signals
over a common medium, particularly
subclasses 58.1+ for time division multiplex (TDM) switching, subclasses
85.1+ for time division bus transmission, and subclasses
91+ for asynchronous TDM communications including addressing.
Pulse or Digital Communications, various subclasses for generic pulse or digital
communication systems and synchronization of clocking signals from
input data.
Telephonic Communications, various subclasses for two-way electrical communication of
intelligible audio information of arbitrary content over a link
including an electrical conductor.
Image Analysis, various subclasses for operations performed on image
data with the aim of measuring a characteristic of an image, detecting
variations, detecting structures, or transforming the image data,
and for procedures for analyzing and categorizing patterns present
in image data.
Education and Demonstration,
subclasses 112+ for communication aids for the handicapped, subclasses
156+ for education and demonstration of language,
subclasses 322+ for question or problem eliciting response.
Telecommunications, appropriate subclasses for modulated carrier wave communication, per
se, and
subclass 26.1 for subject matter which blocks access to a signal
source or otherwise limits usage of modulated carrier equipment.
Data Processing: Generic Control Systems or Specific
Applications,
subclasses 1 through 89for data processing generic control systems, subclasses
90-306 for applications of computers in various environments.
Data Processing: Measuring, Calibrating, or Testing, appropriate subclasses for the application of computer
data processing in measuring, calibrating, or testing.
Electrical Computers: Arithmetic Processing and
Calculating,
subclasses 1+ for hybrid computers, subclasses 100+ for
calculators, digital signal processing and arithmetical processing, per
se, subclasses 300+ for digital filters, and subclasses
800+ for electric analog computers.
Electrical Computers and Digital Processing Systems:
Support,
subclass 187 and 188 for software program protection or computer
virus detection in combination with data encryption.
Error Detection/Correction and Fault
Detection/Recovery, various subclasses for generic electrical pulse
or pulse coded data error detection and correction.
Data Processing: Presentation Processing of Document,
Operator Interface Processing, and Screen Saver Display Processing,
subclasses 243 through 272for document processing including layout, editing,
and spell-checking.
Data Processing: Software Development, Installation,
and Management, appropriate subclasses for significant details of
the construction, analysis, or modification of computer languages.
SECTION V - GLOSSARY
The terms below have been defined for purposes of
classification in this class and are shown in underlined
type when used in the class and subclass definitions.
When these terms are not underlined in the definitions, the meaning
is not restricted to the glossary definitions below.
CORRELATION
A statistical measurement of the interdependence or association
between two variables that are quantitative or qualitative in nature.
A typical calculation would be performed by multiplying a signal
by either another signal (cross-correlation) or by a delayed version
of itself (autocorrelation).
DISTANCE
A statistical measurement for comparing elements defined
by variables or vectors using scalar or vector subtraction of those
elements. Examples: distance=a-b, |a-b|,
(a-b).5 or two vectors may be treated as objects such that the straight
line distance is measured between them.
EXCITATION
Stimulation of the vocal tract by vibratory action of
the vocal cords or by a turbulent air flow. In a digital system,
the vocal tract is typically modelled with a filter and excitation of the filter is performed
using time representations of pitch (voicedexcitation)
and noise (unvoicedexcitation).
LANGUAGE
A systematic means of communicating ideas or feelings by
the use of conventionalized sounds, gestures, or marks having understood
meanings.
LINGUISTICS
The study of human speech including
the units, nature, structure, and modification of language.
Masking
1. The interference with the perception of one sound (the
signal) with another sound (the masker). 2. The number of decibels
by which a masking sound will raise (or change) a listener"s
threshold of audibility of other sounds.
Critical bandwidths
Bandwidths of the hearing process, as measured by the masking
effect of a white, random noise in which a person detects a pure
tone.
Bark spectrum
The width of one critical band.
Mel
A subjective measure of pitch based upon a signal of 1000
Hz. being defined as "1000 mels" where a perceived frequency twice
as high is defined as 2000 mels and half as high as 500 mels.
NOISE
Any sound which is undesirable and interferes with one"s
hearing or with a system"s analysis of desired sound.
Phon
The loudness level of any other sound based upon the SPL
(sound pressure level measured in decibels) of a 1 kHz tone. For
example, if we judge a certain waveform to sound as loud as a 1
kHz tone at 70 dB, then this waveform has a loudness level of 70
phons.
PITCH
The measurable frequency or period at which the glottis vibrates.
SIMILARITY
A statistical measurement which is inversely proportional
to distance. For example, if two
patterns are compared yielding a small distance,
then the patterns would exhibit a large (or high degree of) similarity.
Sone
A measure of loudness as a function of frequency and sound
pressure. A pure tone of 1 kHz. at 40 db above a normal listener"s
threshold produces a loudness of 1 sone.
SPEECH
The communication or expression of thoughts in spoken words.
UNVOICED
Speech sounds produced
by a turbulent flow of air created at some point of stricture in
the vocal tract and usually lacking pitch.
VOICED
Speech sounds produced
by vibratory action of the vocal cords and usually having pitch.
This subclass is indented under the class definition. Subject matter including means or steps for constructing
a word, a phrase, or a sentence in a language.
This subclass is indented under subclass 1. Subject matter wherein a language (i.e.,
source language) stored in a memory
means is translated into another language (i.e.,
target language).
Data Processing: Design and Analysis of Circuit
or Semiconductor Mask,
subclass 3 for translation of computer program in designing
and analyzing circuits and semiconductor mask.
Data Processing: Software Development, Installation,
and Management,
subclasses 136 through 161for software program code translator or compiler
in software development.
This subclass is indented under subclass 2. Subject matter wherein the translation machine includes
a means for reading into the memory means a language,
for pronouncing the translated language or
a particular user interface.
(1)
Note. Examples of such devices include an optical scanner
or voice synthesizer.
This subclass is indented under subclass 2. Subject matter wherein the translation machine includes
a means for providing translation for a specified portion of a sentence
or a clause.
This subclass is indented under subclass 2. Subject matter wherein the translation machine translates
a compound word formed by hyphenation or sentences with quotation marks,
colons, semicolons, or parentheses.
This subclass is indented under subclass 2. Subject matter including a means for assigning storage locations
or accessing addresses to the memory means.
This subclass is indented under subclass 1. Subject matter including means or steps to adapt to, process,
or support plural languages in systems
or in software (i.e., providing language identifiers
on files or providing screen prompts in a selected language),
or to support the conventions or peculiarities of various national languages (i.e., alphabetical ordering, date
or currency indications).
Data Processing: Presentation Processing of Document,
Operator Interface Processing, and Screen Saver Display Processing,
subclasses 264 through 265for composing or editing multiple languages in
a document and subclass 866 for customization or edition of operator
interfaces.
This subclass is indented under subclass 1. Subject matter includes a means for applying grammatical
rules or other analyses (e.g., morphemic, syntax, semantic, etc.)
to define the true meaning of a sentence or phrase.
(1)
Note. When words are undefined in the dictionary of a natural language, the grammatical rules or other
analyses are applied in order to determine the true meaning of a
sentence or a phrase.
Data Processing: Database and File Management,
Data Structures, and Document Processing,
subclasses 1+ , for nominal natural language processing
used in database search and retrieval.
This subclass is indented under subclass 1. Subject matter including a construction, a change, or an
orderly arrangement of dictionary, thesauri, or the like.
Data Processing: Presentation Processing of Document,
Operator Interface Processing, and Screen Saver Display Processing,
subclasses 259 through 260for mere use of a dictionary in editing or composition
of a document.
This subclass is indented under the class definition. Subject matter wherein the system performs operations or
functions on signals which represent speech.
This subclass is indented under subclass 200. Subject matter wherein an operation on the signal is based
upon the masking behavior of the human auditory system.
(1)
Note. The calculation of masking thresholds based upon incoming
analysis of audio is the basis of psychoacoustic compression because
the frequency with the highest local amplitude will tend to mask
(make inaudible) nearby frequencies below the threshold.
(2)
Note. MPEG (Motion Picture Experts Group) sets international
standards such as MPEG 1, level 3 (commonly called MP3) for psychoacoustic
coding to achieve audio compression of up to 10:1. Typical coders
work on a 16-bit PCM audio signal, which is the typical CD quality
standard.
(3)
Note. Only white noise in a bandwidth centered about a tone
and less than or equal to the critical bandwidth contributes to
the masking effect. Critical bands are generally considered a set
of filters or channels tuned to different center frequencies having
a bandwidth of less than a third of an octave.
(4)
Note. A plot of frequency versus pitch in mels is similar
in shape to the plot of frequency versus the position of auditory-nerve
patches on the basilar membrane. This is evidence that human judgment
of pitch is based upon the point of excitation along the basilar membrane
in the ear.
This subclass is indented under subclass 201. Subject matter wherein coding is performed using parallel
distributed processing elements constructed in hardware or simulated
in software.
This subclass is indented under subclass 201. Subject matter wherein the speech is
encoded using a specific mathematical function (e.g., Fourier, Walsh,
cosine/sine transform, etc.).
This subclass is indented under subclass 203. Subject matter wherein the function is orthogonal (transformations
as applied to vector, matrix, linear and polynomial functions, for example).
This subclass is indented under subclass 206. Subject matter wherein the specific speech information
represents the predominant frequency of the speech.
This subclass is indented under subclass 207. Subject matter wherein the specific speech information
represents the presence (voiced)
or absence (unvoiced) of predominant
frequency components.
This subclass is indented under subclass 206. Subject matter wherein the specific speech information
represents the frequency values of any of several resonance bands
which determine the phonetic quality of a vowel sound.
This subclass is indented under subclass 201. Subject matter wherein the speech signal
is represented using time (e.g., time measurements and energy measured
over time).
This subclass is indented under subclass 211. Subject matter wherein the signal is sampled over time,
and the magnitude of each sample is quantized and converted into
a digital signal.
This subclass is indented under subclass 211. Subject matter wherein time measurements are used to determine
the presence (voiced) or absence
(unvoiced) of predominant frequency components.
This subclass is indented under subclass 211. Subject matter wherein time measurements are used to determine
the presence or absence of speech (e.g.,
pauses between words, etc.).
This subclass is indented under subclass 216. Subject matter wherein the relationships are between different speech samples taken from the same time
series.
This subclass is indented under subclass 201. Subject matter wherein the speech signal
is coded and corrected by the difference of the decoded coded signal
from the original speech signal.
This subclass is indented under subclass 221. Subject matter wherein the encoding maps a sequence of continuous
or discrete vectors into a digital sequence.
This subclass is indented under subclass 221. Subject matter wherein the encoding models speech using
representations including the primary frequency period or periods
(e.g., pitchexcitation,
multipulse excitation, etc.).
This subclass is indented under subclass 201. Subject matter wherein modifications of the speech signal
emphasize or deemphasize certain features (e.g., spectral slope,
average power, etc.).
This subclass is indented under subclass 226. Subject matter wherein decoding after transmission minimizes
the effects of noise in the transmission
path.
This subclass is indented under subclass 201. Subject matter wherein limited storage or transmission resources
are allocated by giving more resources to areas containing more
data and giving fewer resources to areas containing less data.
This subclass is indented under subclass 200. Subject matter wherein speech is
separated into discrete components which are distinguished from
one another.
This subclass is indented under subclass 231. Subject matter using parallel distributed processing elements
constructed in hardware or simulated in software.
This subclass is indented under subclass 231. Subject matter wherein the discrete components are modified
to emphasize or deemphasize certain features (e.g., spectral slope, average
power, etc.).
This subclass is indented under subclass 231. Subject matter wherein the distinguished discrete components
are converted into image output (e.g., text).
This subclass is indented under subclass 236. Subject matter wherein the specific function measures a correlation between discrete components
(e.g., absolute magnitude difference functions (AMDF), autocorrelation, cross-correlation,
etc.).
This subclass is indented under subclass 236. Subject matter wherein the specific function uses probability
to determine the occurrence of a discrete component.
This subclass is indented under subclass 236. Subject matter wherein time components of the discrete components
are aligned with reference components (e.g., using dynamic programming).
This subclass is indented under subclass 236. Subject matter wherein discrete components are distinguished
by traversing possible paths through a time series.
This subclass is indented under subclass 231. Subject matter including specific methods for registering
the discrete components to be used as references.
This subclass is indented under subclass 243. Subject matter wherein similar references are placed or
divided into groups (e.g., K-means algorithm, nearest neighbor,
etc.).
This subclass is indented under subclass 231. Subject matter wherein different voices are distinguished
(e.g., speaker identification or verification).
This subclass is indented under subclass 246. Subject matter including separating speech into sound
segments (e.g., utterances, words, phonemes, allophones, etc.).
This subclass is indented under subclass 251. Subject matter including models which describe the interconnections
between words or subportions of words.
This subclass is indented under subclass 255. Subject matter wherein the models include states which represent speech sound portions and transitions
which represent connections between speech sound
portions (e.g., hidden Markov models, heuristic Markov models, etc.).
This subclass is indented under subclass 256. Subject matter wherein a Markov chain used in the recognition
process has un-observable (hidden) states.
(1)
Note. The subject matter in this subclass is substantially
the same in scope as ECLA (G10L 15/14M).
(2)
Note. The observation model itself is part of the stochastic
process (Markov Chain) with an underlying stochastic process that
is not directly observable, but can be observed through a set of
stochastic processes that produce the sequence of observations.
(3)
Note. The HMM has different elements, including the following –
number of states, the number of distinct observations per state,
state transition probability distribution, the observation symbol probability
distribution, and the initial state distribution.
(4)
Note. The manipulation of HMM s can be use in improving the
probability of observation sequences, optimizing state sequences,
or maximizing the probability of the state sequences.
(5)
Note. Subcategories to the types of HMM s include finite
state, discrete versus continuous, mixture densities, autoregressive,
null transition, tied states, and state duration.
This subclass is indented under subclass 256.1. Subject matter wherein the models include a learning
process for recognizing speech data, e.g., the construction of a
library of models for the words in a vocabulary, including the states.
(1)
Note. The subject matter in this subclass is substantially
the same in scope as ECLA (G10L 15/14M1).
This subclass is indented under subclass 256.2. Subject matter wherein intrinsic parameters of the HMM are
modified to overcome lack of training data, and to simplify the
model, e.g., state sharing, tying, and deleted interpolation.
(1)
Note. The subject matter in this subclass is substantially
the same in scope as ECLA (G10L 15/14M1S).
(2)
Note. State sharing involves combining two or more separately
trained models, one of which is more reliably trained than the other.
The scenario in which this can happen is the case when we use tied states
which forces "different" states to share an identical
statistical characterization, effectively reducing the number of parameters
in the model.
(3)
Note. Parameter tying involves setting up an equivalence
relation between HMM parameters in different states. In this manner
the number of independent parameters in the model is reduced and the
parameter estimation becomes somewhat simpler and in some cases
more reliable. Parameter tying is used when the observation density,
for example, is known to be the same in two or more states.
(4)
Note. Deleted interpolation is a parameter method aimed to
improve model reliability. The concept involves combining two or
more separately trained models, one of which is more reliably trained than
the other. The scenario in which this can happen is the case when
we use tied states which forces "different" states
to share an identical statistical characterization, effectively
reducing the number of parameters in the model. The technique of
deleted interpolation has been successfully applied to a number
of problems in speech recognition, including the estimation of trigram
word probabilities for language models, and the estimation of HMM
output probabilities for trigram phone models.
This subclass is indented under subclass 256.1. Subject matter wherein the HMM includes a duration state
model for speech recognition, e.g., semi HMM’s segmental
models, and transition probabilities.
(1)
Note. The subject matter in this subclass is substantially
the same in scope as ECLA (G10L 15/14M2).
(2)
Note. A semi- Markov HMM is like an HMM except each state
can emit a sequence of observations.
(3)
Note. Within a state segment models introduce dependency
between frames via their common dependence on a trajectory. There
may be only a single trajectory or a continuous mixture of trajectories.
The probability distribution over the sequence of frames for a state, given
the duration and trajectory, is then typically modeled as independent
Gaussian distributions for each time step, centered on the trajectory.
(4)
Note. Symbol emission probabilities are associated to the
states and transition probabilities to the connections between them.
This subclass is indented under subclass 256.1. Subject matter including a HMM structure wherein subgroups
of HMM types are used to perform speech recognition.
(1)
Note. The subject matter in this subclass is substantially
the same in scope as ECLA (G10L 15/14M3).
(2)
Note. Each subgroup can vary by type of model, model size,
and observation symbols.
This subclass is indented under subclass 256.1. Subject matter wherein the HMM contains probability density
function such that an emission probability is calculated for each
state within the model.
(1)
Note. The subject matter in this subclass is substantially
the same in scope as ECLA (G10L 15/14M4).
(2)
Note. For each state j,
and for each possible output, a probability that a particular output
symbol o is observed in that state.
This is represented by the function bj(o), which
gives the probability that o is
emitted in state j. This is called the
emission probability.
This subclass is indented under subclass 256.6. Subject matter wherein the HMM contains continuous probability
density observation models for the purpose of avoiding possible signal
degradation inherent with discrete representations of signals.
(1)
Note. The subject matter in this subclass is substantially
the same in scope as ECLA (G10L 15/14M4C).
This subclass is indented under subclass 256.6. Subject matter wherein the HMM contains discrete probability
density observation models which allows for the use of a discrete
probability density within each state of the model.
(1)
Note. The subject matter in this subclass is substantially
the same in scope as ECLA (G10L 15/14M4D).
(2)
Note. Discrete probability density is used when the state
of the model is discrete (e.g. representing a letter of the alphabet).
Vector quantization is used to model its state.
This subclass is indented under subclass 200. Subject matter wherein component parts of a speech signal
are combined to produce a synthetic speech output.
This subclass is indented under subclass 258. Subject matter wherein synthetic speech output is
formed using parallel distributed processing elements constructed
in hardware or simulated in software.
This subclass is indented under subclass 258. Subject matter wherein the component parts are represented
by coefficients derived from a sequence of past speech samples.
This subclass is indented under subclass 258. Subject matter wherein the component parts are represented
by coefficients derived from relationships between time series speech samples.
This subclass is indented under subclass 258. Subject matter wherein the component parts are represented
by the period of the primary frequency of the speech signal
(e.g., pitchexcitation,
multi-pulse excitation, etc.).
This subclass is indented under subclass 258. Subject matter wherein the component parts are combined
using estimates of intermediate values (e.g., waveform smoothing).
This subclass is indented under subclass 258. Subject matter wherein the component parts are combined
or linked together in a defined manner (e.g., Markov models, trees,
tries (tables representing trees), graphs, etc.).
This subclass is indented under subclass 258. Subject matter wherein the component parts comprise time
based elements (e.g., words, phonemes, allophones, etc.).
This subclass is indented under subclass 258. Subject matter wherein the component parts comprise frequency
based elements (e.g., pitch variations,
inflection, formants, etc.).
This subclass is indented under subclass 258. Subject matter wherein the component parts are restored
to speech using specific mathematical
functions (e.g., Fourier, Walsh, Hilbert, Z-transform, cosine/sine
transforms, etc.).
This subclass is indented under subclass 200. Subject matter intended or designed for a specified use
to which the speech signal processing is
being applied.
This subclass is indented under subclass 270. Subject matter wherein a system that employs speech recognition
or synthesis to control or to provide user feedback such that the
processing of speech data may occur at various levels within a computer
network.
(1)
Note. Various levels of processing would include local or
remote locations relative to the user in order to make use of available
resources. For example, a local terminal might not have the necessary
storage or processing power but this can be overcome by accessing
resources over a network. Such resources may include the raw processing
power necessary for analysis and pattern matching as well as dictionaries
having data relevant to large vocabularies and multiple languages.
(2)
Note. Nominal recitations of speech or audio in network applications
are classified elsewhere.
Multiplex Communications,
subclasses 229 through 240for data flow congestion prevention or control,
subclasses 260-269 for conferencing and subclass 351 for voice over
internet.
Electrical Computers and Digital processing systems:
Multiple computer or Process Coordinating,
subclasses 227 through 229for network computer-to-computer connections.
This subclass is indented under subclass 270. Subject matter for assisting handicapped people (e.g., blind
or speech impaired communication
and control).
This subclass is indented under subclass 270. Subject matter wherein speech is
edited using waveform portions or other representations of the sounds
to be modified.
This subclass is indented under the class definition. Subject matter where there is either an expansion or reduction
of the bandwidth required for transmission of a sound signal.
(1)
Note. This subclass and its indents provide for bandwidth
compression or expansion of audio signals other than speech signals.
Pulse or Digital Communications,
subclasses 240 through 241for bandwidth compression or expansion of a pulse
or digital signal, particularly subclasses 240.01-240.29 for digital television.
This subclass is indented under subclass 500. Subject matter combined with means to discard and replace
redundant information by a code indicating what has been discarded.
AUDIO SIGNAL TIME COMPRESSION OR EXPANSION (E.G., RUN LENGTH CODING):
This subclass is indented under the class definition. Subject matter where there is either an expansion or reduction
of the time required for transmission of a nonspeech sound
signal.
This subclass is indented under subclass 503. Subject matter combined with means to discard and replace
redundant information by a code indicating what has been discarded.
Coded Data Generation or Conversion,
subclass 55 for content reduction encoding, per se.
E-SUBCLASSES
The E-subclasses in U.S. Class 704 provide for methods and
devices for analyzing or synthesizing spoken language and for detecting,
recognizing, or modifying speech signal characteristics.
MISCELLANEOUS ANALYSIS OR DETECTION OF SPEECH CHARACTERISTICS
(EPO):
This main group provides for processes and apparatus for
analyzing or detecting speech characteristics not provided for elsewhere.
This subclass is substantially the same in scope as ECLA classification
G10L11/00.
This main group provides for processes and apparatus for
synthesizing speech. This subclass is substantially the same in
scope as ECLA classification G10L13/00.
This main group provides for processes, systems, and apparatus
for the recognition of speech, including training of speech recognition
systems, language recognition, speech classification and search,
speech-to-text systems, and evaluation or assessment of speech recognition
systems. This subclass is substantially the same in scope as ECLA
classification G10L15/00.
This main group provides for processes and apparatus for
recognizing special voice characteristics, systems using speaker
recognizers and details of speaker identification or verification
processes or apparatus. This subclass is substantially the same
in scope as ECLA classification G10L17/00.
SPEECH OR AUDIO SIGNAL ANALYSIS-SYNTHESIS TECHNIQUES FOR REDUNDANCY
REDUCTION, E.G., IN VOCODERS, ETC.; CODING OR DECODING OF SPEECH
OR AUDIO SIGNALS; COMPRESSION OR EXPANSION OF SPEECH OR AUDIO SIGNALS,
E.G., SOURCE-FILTER MODELS, PSYCHOACOUSTIC ANALYSIS, ETC. (EPO):
This main group provides for processes and apparatus for
the coding, decoding, compression or expansion of speech or audio
signals, including techniques for redundancy reduction, and psychoacoustic
analysis. This subclass is substantially the same in scope as ECLA
classification G10L19/00.
MODIFICATION OF AT LEAST ONE CHARACTERISTIC OF SPEECH WAVES (EPO):
This main group provides for processes and apparatus for
modifying at least one characteristic of a speech signal. This subclass
is substantially the same in scope as ECLA classification G10L21/00.
Note: The Patent
and Trademark Depository Library Program (PTDLP) administers a nationwide
network of public, state and academic libraries designated as Patent and
Trademark Depository Libraries authorized by 35 U.S.C. 13 to: Disseminate
Patent and Trademark Information Support Diverse Intellectual Property
Needs of the Public. PTDL Contact Information
Note: For information/comments on electronic information products, such
as purchasing USPTO data, or to discuss system requirements for magnetic
tape products, contact:
KEY: =online
business system =fees =forms=help =laws/regulations =definition
(glossary)
The Inventors
Assistance Center is available to help you on patent
matters.Send questions about USPTO programs and services
to theUSPTO
Contact Center (UCC). You can suggest USPTO webpages
or material you would like featured on this section by E-mail
to the webmaster@uspto.gov.
While we cannot promise to accommodate all requests, your suggestions
will be considered and may lead to other improvements on the
website.