CN110223688A - A kind of self-evaluating system of compressed sensing based hepatolenticular degeneration disfluency - Google Patents
A kind of self-evaluating system of compressed sensing based hepatolenticular degeneration disfluency Download PDFInfo
- Publication number
- CN110223688A CN110223688A CN201910494049.XA CN201910494049A CN110223688A CN 110223688 A CN110223688 A CN 110223688A CN 201910494049 A CN201910494049 A CN 201910494049A CN 110223688 A CN110223688 A CN 110223688A
- Authority
- CN
- China
- Prior art keywords
- disfluency
- illness
- speech
- patient
- voice
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 208000002972 Hepatolenticular Degeneration Diseases 0.000 title claims abstract description 20
- 238000000034 method Methods 0.000 claims abstract description 44
- 201000008482 osteoarthritis Diseases 0.000 claims abstract description 44
- 238000011156 evaluation Methods 0.000 claims abstract description 38
- 238000012360 testing method Methods 0.000 claims abstract description 12
- 238000004458 analytical method Methods 0.000 claims description 21
- 238000001228 spectrum Methods 0.000 claims description 17
- 238000004422 calculation algorithm Methods 0.000 claims description 13
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims description 13
- 238000000605 extraction Methods 0.000 claims description 13
- 201000010099 disease Diseases 0.000 claims description 11
- 238000002474 experimental method Methods 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 10
- 208000012902 Nervous system disease Diseases 0.000 claims description 7
- 239000000203 mixture Substances 0.000 claims description 7
- 208000011580 syndromic disease Diseases 0.000 claims description 7
- 206010013887 Dysarthria Diseases 0.000 claims description 5
- 150000001875 compounds Chemical class 0.000 claims description 5
- 238000009472 formulation Methods 0.000 claims description 4
- 230000033764 rhythmic process Effects 0.000 claims description 4
- 238000003860 storage Methods 0.000 claims description 4
- 230000000153 supplemental effect Effects 0.000 claims description 4
- 238000000926 separation method Methods 0.000 claims description 3
- 230000015572 biosynthetic process Effects 0.000 claims description 2
- 238000003786 synthesis reaction Methods 0.000 claims 1
- 238000005070 sampling Methods 0.000 abstract description 9
- 208000011293 voice disease Diseases 0.000 abstract description 7
- 238000013144 data compression Methods 0.000 abstract description 2
- 238000011160 research Methods 0.000 description 26
- 230000006870 function Effects 0.000 description 22
- 238000005516 engineering process Methods 0.000 description 16
- 238000009432 framing Methods 0.000 description 13
- 238000011161 development Methods 0.000 description 11
- 230000033001 locomotion Effects 0.000 description 10
- 230000004888 barrier function Effects 0.000 description 9
- 239000003814 drug Substances 0.000 description 7
- 208000024891 symptom Diseases 0.000 description 7
- 238000007906 compression Methods 0.000 description 6
- 230000006835 compression Effects 0.000 description 6
- 238000001914 filtration Methods 0.000 description 6
- 239000011159 matrix material Substances 0.000 description 6
- 210000003205 muscle Anatomy 0.000 description 6
- 210000000056 organ Anatomy 0.000 description 6
- 238000012545 processing Methods 0.000 description 6
- 208000005392 Spasm Diseases 0.000 description 5
- 210000004227 basal ganglia Anatomy 0.000 description 5
- 238000001514 detection method Methods 0.000 description 5
- 238000012549 training Methods 0.000 description 5
- RYGMFSIKBFXOCR-UHFFFAOYSA-N Copper Chemical compound [Cu] RYGMFSIKBFXOCR-UHFFFAOYSA-N 0.000 description 4
- 230000008859 change Effects 0.000 description 4
- 229910052802 copper Inorganic materials 0.000 description 4
- 239000010949 copper Substances 0.000 description 4
- 230000006378 damage Effects 0.000 description 4
- 238000003745 diagnosis Methods 0.000 description 4
- 230000006872 improvement Effects 0.000 description 4
- 230000003902 lesion Effects 0.000 description 4
- 210000001103 thalamus Anatomy 0.000 description 4
- 230000001755 vocal effect Effects 0.000 description 4
- 208000012661 Dyskinesia Diseases 0.000 description 3
- 208000027776 Extrapyramidal disease Diseases 0.000 description 3
- 206010044565 Tremor Diseases 0.000 description 3
- 210000001638 cerebellum Anatomy 0.000 description 3
- 210000004087 cornea Anatomy 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 210000000653 nervous system Anatomy 0.000 description 3
- 238000012827 research and development Methods 0.000 description 3
- 206010003591 Ataxia Diseases 0.000 description 2
- 241001672694 Citrus reticulata Species 0.000 description 2
- 206010011878 Deafness Diseases 0.000 description 2
- 206010013931 Dyslalia Diseases 0.000 description 2
- 206010016654 Fibrosis Diseases 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000005311 autocorrelation function Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000033228 biological regulation Effects 0.000 description 2
- 210000004556 brain Anatomy 0.000 description 2
- 210000001159 caudate nucleus Anatomy 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 230000007882 cirrhosis Effects 0.000 description 2
- 208000019425 cirrhosis of liver Diseases 0.000 description 2
- 210000001653 corpus striatum Anatomy 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 208000035475 disorder Diseases 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 210000000214 mouth Anatomy 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000008447 perception Effects 0.000 description 2
- 238000007637 random forest analysis Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 210000002966 serum Anatomy 0.000 description 2
- 238000010183 spectrum analysis Methods 0.000 description 2
- 208000027765 speech disease Diseases 0.000 description 2
- 238000002630 speech therapy Methods 0.000 description 2
- 210000001260 vocal cord Anatomy 0.000 description 2
- 101150075644 ATP7B gene Proteins 0.000 description 1
- 208000009017 Athetosis Diseases 0.000 description 1
- 208000022306 Cerebral injury Diseases 0.000 description 1
- 108010075016 Ceruloplasmin Proteins 0.000 description 1
- 102100023321 Ceruloplasmin Human genes 0.000 description 1
- 206010010947 Coordination abnormal Diseases 0.000 description 1
- 206010011224 Cough Diseases 0.000 description 1
- JPVYNHNXODAKFH-UHFFFAOYSA-N Cu2+ Chemical compound [Cu+2] JPVYNHNXODAKFH-UHFFFAOYSA-N 0.000 description 1
- 208000019505 Deglutition disease Diseases 0.000 description 1
- 206010064571 Gene mutation Diseases 0.000 description 1
- 206010067125 Liver injury Diseases 0.000 description 1
- 208000026072 Motor neurone disease Diseases 0.000 description 1
- 208000002038 Muscle Hypertonia Diseases 0.000 description 1
- 208000002740 Muscle Rigidity Diseases 0.000 description 1
- 208000018737 Parkinson disease Diseases 0.000 description 1
- 206010039424 Salivary hypersecretion Diseases 0.000 description 1
- 208000003028 Stuttering Diseases 0.000 description 1
- 208000018839 Wilson disease Diseases 0.000 description 1
- 210000001015 abdomen Anatomy 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 208000025341 autosomal recessive disease Diseases 0.000 description 1
- 238000007630 basic procedure Methods 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 230000006931 brain damage Effects 0.000 description 1
- 231100000874 brain damage Toxicity 0.000 description 1
- 208000029028 brain injury Diseases 0.000 description 1
- 230000010336 brain pathway Effects 0.000 description 1
- 238000000180 cavity ring-down spectroscopy Methods 0.000 description 1
- 210000003591 cerebellar nuclei Anatomy 0.000 description 1
- 210000003710 cerebral cortex Anatomy 0.000 description 1
- 230000002490 cerebral effect Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 210000000038 chest Anatomy 0.000 description 1
- MYSWGUAQZAJSOK-UHFFFAOYSA-N ciprofloxacin Chemical compound C12=CC(N3CCNCC3)=C(F)C=C2C(=O)C(C(=O)O)=CN1C1CC1 MYSWGUAQZAJSOK-UHFFFAOYSA-N 0.000 description 1
- 238000003759 clinical diagnosis Methods 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 239000000039 congener Substances 0.000 description 1
- 229910001431 copper ion Inorganic materials 0.000 description 1
- 210000003792 cranial nerve Anatomy 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 230000036425 denaturation Effects 0.000 description 1
- 238000004925 denaturation Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000008021 deposition Effects 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 230000000142 dyskinetic effect Effects 0.000 description 1
- 238000013399 early diagnosis Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 230000008451 emotion Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000002360 explosive Substances 0.000 description 1
- 210000001905 globus pallidus Anatomy 0.000 description 1
- 210000003128 head Anatomy 0.000 description 1
- 231100000234 hepatic damage Toxicity 0.000 description 1
- 208000016290 incoordination Diseases 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 230000002401 inhibitory effect Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 230000003834 intracellular effect Effects 0.000 description 1
- 238000007917 intracranial administration Methods 0.000 description 1
- 230000003907 kidney function Effects 0.000 description 1
- 210000004185 liver Anatomy 0.000 description 1
- 230000008818 liver damage Effects 0.000 description 1
- 208000019423 liver disease Diseases 0.000 description 1
- 230000003908 liver function Effects 0.000 description 1
- 210000004072 lung Anatomy 0.000 description 1
- 230000003340 mental effect Effects 0.000 description 1
- 210000001259 mesencephalon Anatomy 0.000 description 1
- 230000004060 metabolic process Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 208000005264 motor neuron disease Diseases 0.000 description 1
- 239000002858 neurotransmitter agent Substances 0.000 description 1
- 230000000474 nursing effect Effects 0.000 description 1
- 230000037361 pathway Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 208000027232 peripheral nervous system disease Diseases 0.000 description 1
- 210000002975 pon Anatomy 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000002203 pretreatment Methods 0.000 description 1
- 238000004393 prognosis Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000011155 quantitative monitoring Methods 0.000 description 1
- 230000000241 respiratory effect Effects 0.000 description 1
- 230000029058 respiratory gaseous exchange Effects 0.000 description 1
- 208000026451 salivation Diseases 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000000542 thalamic effect Effects 0.000 description 1
- 230000001225 therapeutic effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
- 230000032895 transmembrane transport Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/08—Speech classification or search
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
- G10L25/18—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being spectral information of each sub-band
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
- G10L25/24—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being the cepstrum
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/45—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of analysis window
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
- G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The present invention relates to a kind of self-evaluating systems of compressed sensing based hepatolenticular degeneration disfluency, it is characterised in that: including the automatic evaluation module of dysarthrosis, the automatic evaluation module of voice disorder, state of an illness classification automatic identification module and the automatic evaluation module of state of an illness grade.The present invention proposes the compressed sensing sampling of improved voice signal, strong hardware supported and data compression technique are provided to remote patient data sampling, for realizing that the remote collection of data is of great immediate significance with test, acquisition for each area sample data needed for the building of later data library lays the foundation, and also provides powerful technical support to the use of all patients for automatic identification system.
Description
Technical field
The present invention relates to a kind of self-evaluating systems of compressed sensing based hepatolenticular degeneration disfluency.
Background technique
Hepatolenticular degeneration (HLD) is also known as Wilson sick (Wilson ' s disease, WD), is that a kind of copper dysbolism is led
The autosomal recessive hereditary diseases for causing brain Basal ganglia denaturation and hepatic disorder, is apt to occur in teenager, because positioned at 13q14.3's
ATP7B gene mutation causes intracellular copper ion transmembrane transport obstacle, so that copper is at positions such as liver, cerebral basal ganglia area and corneas
Deposition[1].Lenticular nucleus is especially invaded, it is earliest and obvious with shell core, it is secondly globus pallidus, caudate nucleus and cerebral cortex, thalamus bottom
Core, rubrum, black substance, thalamus and dentate nucleus can also be involved.Clinical manifestation progressive aggravate extrapyramidal symptom, cornea K-F ring,
Cirrhosis, mental symptom and kidney function damage etc.[2].The complicated clinical manifestation multiplicity of its extrapyramidal symptom, is mainly shown as structure
Sound barrier hinders (dysarthria), expression exception, salivates, is clumsy in one's movement, lurch, tremble, splinting, dancing, athetoid
Deng.Wherein dysarthrosis is one of main performance[3], which reduce patient's speech articulation and ability to exchange, influence the life of patient
Bioplasm amount and social-life ability[4].Rare in most of American-European countries WD, disease incidence is (15-30)/1,000,000.Though China
Report on Epidemiological without large data, it is generally understood that disease incidence is more much higher than western countries.Early stage dysarthrosis is easy
It is ignored, it is especially the WD of onset symptoms often by sing misdiagnosis and mistreatment using slight dysarthrosis[5], so that missing best occasion for the treatment.WD
Dysarthrosis performance is varied, predominantly tongue, oral cavity and the structures sound organ such as pharyngeal because hypermyotonia and (or) tremble etc. because
Dyslalia caused by element and (or) cannot.Less serious case show as speaking the rhythm and pace of moving things is slow, tone is droning, voice trembling, harmonious sounds disorder,
Language is not smooth, but not appreciably affects exchange；It is ambiguous to enunciate compared with severe one, and hardly possible is understood by other people, can be exchanged reluctantly with short sentence,
Influence exchange；Serious person then completely loses oral communication ability, cannot exchange with people[6,7]。
WD dysarthrosis parting is divided into 6 seed types according to nervous system damage position and the speech extent of damage, i.e., spasm type,
Relaxation type moves reduction type, is moved through many types of, disorder type and mixed type[6,8,9].Dysarthrosis type and severity and brain damage
Harmful position and severity is related[10].The dysarthrosis the above-mentioned type performance of WD exists[6,11,12], visible in iconography to deposit
Intracranial lesion, which is read, in the government official of dysarthrosis is concentrated mainly on the portions such as lenticular nucleus, midbrain, pons, head of caudate nucleus, thalamus, cerebellum
Position[13].Clinical visible spasm type is spoken slowly arduously, and tone is low and single, is often accompanied by dysphagia；Relaxation type is spoken with nose
Sound, consonant, vowel express mistake, often exist and feed cough of choking；Incoordination type structure sound is inaccurate, in explosive voice, phoneme delays,
It pauses, extends in sample of reciting a poem；Be moved through it is many types of speak it is sometimes fast and sometimes slow, vowel extend, modified tone, volume change excessively or sound in
Only；It is low flat to move reduction type pronunciation, monotone, or even have trill and stutter, it moves inappropriate with salivation；Mixed type part table
Now it is spasm type and relaxation type dysarthrosis, is to write with spasm type ingredient, it is more common for the very few former spasm of ataxia original movement
Type[6,9,14].WD patient is common with mixed type dysarthrosis.
Resume speed is often relatively slow after the treatment of WD dysarthrosis, and the sequelae shape that Chang Liuyou is different degrees of, to patient and
Family members bring psychology and spiritual pressure, constrain comprehensive rehabilitation of patient.The complicated clinical manifestation multiplicity of WD dysarthrosis,
Its corresponding treatment measures and prognosis also have different from dysarthrosis caused by other reasons, therefore accomplish early diagnosis, morning as far as possible
Treatment.All clinics encounter dysarthrosis caused by having unknown cause, especially with cirrhosis or other extrapyramidal symptoms, or
Once there is transient liver damage symptom and had similar family history patient, and to consider that WD is possible, it is raw need to further to improve serum copper
Change (serum copper, ceruloplasmin, CuO-2 layer), liver function, brain MRI and cornea K-F ring etc. check, reduce to the full extent
Mistaken diagnosis is failed to pinpoint a disease in diagnosis.
The treatment modern medicine study of disfluency lays particular emphasis on disfluency evaluation and speech therapy, for there are speech barriers
The WD patient hindered will more pay attention in the specification evaluation for the parting and severity for exploring its disfluency to disfluency symptom
Treatment, catch the opportune moment of early stage speech rehabilitation, in conjunction with disfluency type and severity, take individuation, Chinese medicine
The comprehensive therapeutic plan that treatment is combined with modern medicine Rehabilitation Training in Treating technology, traditional Chinese and western medicine has complementary advantages, and encourage patient
Adhere to that reinforce speech rehabilitation improves the quality of living to can preferably restore its speech function for a long time.
How research organically combines the assessment of speech recognition technology and disfluency and rehabilitation, and artificial intelligence application is arrived
In clinical medicine, realizes computer assisted automation disfluency assessment and rehabilitation, can not only assist diagnosis, control
It treats, it can also be used to which patient voluntarily diagnoses, and for the Finding case state of an illness and the development of the grasp state of an illness in time has great help, therefore
The present invention has very high value for clinical application, has very much social value and realistic meaning.
2 domestic and international present Research analyses:
The research of 2.1WD dysarthrosis；
Early in 1912, Wilson was for the first time described in detail WD, and there are structure sound barriers for majority in 12 patients of report
Hinder, and thinks that slight dysarthrosis is earliest one of nervous system signs[15].The scholars such as later period Fister, Martin have
Similar report[11], domestic Zhu Yun wave etc.[16]Also have and reported by the WD of onset symptoms of dysarthrosis, but not to dysarthrosis
It is elaborated.Most literature describes WD dysarthrosis and exists simultaneously with dyskinesia, Behavioral disturbances, and thinks structure sound barrier
Hinder incidence quite high, the report such as Liu Wei is up to 77.4%[17], with external Machado[18]It reports similar.Related WD dysarthrosis is also
Further quantity table method, instrumental method is needed to carry out analysis and assessment, and need to be groped a set of more effective speech therapy method.
Dysarthrosis is since maincenter or Peripheral nervous system diseases become cause speech related muscles and benumb or move uncoordinated draw
It rises, the lesion from brain pathways to muscle in itself can lead to, and be clinically common aphasis[14], most of structure sound barrier
Hinder and belongs to dyskinesia scope, therefore also known as dysarthria.WD dysarthrosis is mostly dyskinetic structure sound barrier
Hinder[19].Language is a more complicated process, is generally made of the Surrounding muscles and non-musculature of lung and thorax abdomen
Respiratory movement system, cartilagines larynigs and muscle (including vocal cords) composition sonification system and articulatory system three coordinated
At[6].IX, X, XII 3 dominates above-mentioned vocal organs to cranial nerve jointly, and suitably changing by air-flow makes oral cavity issue fluency
Speech, such complex process need precise motion planning and regulation[10,14].WD dysarthrosis occurs and basal ganglion function
It lacks of proper care related, basal ganglion usually makes indirect pathway and direct path function keeps balance, Reasonable Regulation And Control movement, when its function
Then there is dyskinesia in imbalance, causes muscle rigidity not flexible, articulation obstacle, and volume, word speed, the rhythm and pace of moving things change, and causes speech
It is unclear, voice mistake；Some patientss are then that can not send out sound since throat muscle or vocal cords myokinesis are uncoordinated；Lesions on cerebellum can
Cause ataxiophemia, in sample language etc. of reciting a poem；Active speech is reduced after thalamic lesion, and wave volume is substantially reduced with clear
Degree decline etc.[6].To sum up, the WD so breath such as the dysarthrosis of multiplicity and its diseased region basal ganglion, cerebellum, thalamus manner of breathing
Close, the damage of above structure directly or indirectly affects the conduction and metabolism of neurotransmitter, cause secondary sound muscle and
The myodystony of congener, leads to dyslalia[6,14]。
The evaluation of dysarthrosis, the current country there is no unified assessment method[20], there are no special assessment marks for WD dysarthrosis
Standard, majority use Frenchay dysarthrosis evaluation assessment[21]Or improved method and China Rehabilitation Research Center's dysarthrosis check table,
It is checked by clinician or rehabilitation department doctor, scoring, record, evaluation dysarthrosis degree, type[6,21,22]。
China's speech structure sound, voice disorder patient populations are more, and relevant obstacle appraisal procedure is mainly with the subjective sense of hearing
Based on perception, lack certain objectivity and stability.In recent years, speech recognition technology has obtained widely answering in multiple fields
With the application study in terms of verbal language education also achieves certain achievement.But in disfluency assessment and rehabilitation research
Field, the related research result based on speech recognition is actually rare, and fails to cause enough attention.This research is according to the country
The study on assessing method status and development trend of outer speech structure sound, voice disorder, integrated voice identification technology are taught in verbal language
The research achievement of middle application is educated, emphasis is directed to the pilot study that the WD patient with disfluency has carried out automatic assessment, this
It studies and subsequent may extend to commenting for disfluency caused by nervous system other diseases (such as Parkinson, motor neuron disease etc.)
It surveys and diagnoses, during research, we also sample and analyze disfluency patient caused by above several the nervous system diseases
Voice, for the later period parting and automatic judgment lay the foundation.
2.2, the study on assessing method of speech structure sound:
In recent years, with computer technology, signal processing technology, pattern-recognition and the fast development of acoustic technique, language
Sound identification is widely applied in many aspects such as industry, military affairs, traffic, civilian.In verbal language education sector, voice is known
Other application study also achieves certain achievement, by taking the country as an example, be concentrated mainly on foreign language learning, normal child's language education,
Mandarin Grade Exam System etc..But for the disfluency assessment and rehabilitation using voice as main study subject
Educational research field, the related research result based on speech recognition is considerably less, and does not cause enough attention.
External speech science man and phonetician to the research of structure sound organ movement's situation mainly use subjective observation and
Medical image means.The structure sound method of estimating motion of subjective observation can only be provided due to lacking the quantitative monitoring to structure system for electrical teaching
The structures sound organ such as lower song, lip, tongue is qualitatively described, and the professional knowledge of appraiser is required relatively high.Based on medicine
The correlative study of iconography is more early in foreign countries' starting, and the Europe before and after year successively establishes several speech sciences and phonetics
Research center, wherein Germany scientist Menzerath is based on x-ray imaging and makes video recording, and realizes the individual structure of careful observation earlier
The motion conditions of sound organ and the variation of relative position[23], speech science man of the same period also utilizes medical image means to analyze
The structures sound feature [24] such as clear unique, vowel tongue position position of consonant out.Nineteen sixty -1980 years, electronic computer technology
Rapid development, speech science man and phoneticians are no longer satisfied with simple electro-physiological signals class to the research that structure sound moves
Than the channel model of various Computer Simulations comes into being, wherein the most well-known is the P.Mermelstain (1973) in the U.S.
The structure sound model [25] that the speech of proposition generates, the model are that a kind of two-dimentional sound channel is depicted with straight line, curve and angle is imitative
True mode, the assessment which moves structure sound have a very strong directive significance, but due to current nmr imaging technique at
This is high, in actual clinical diagnosis and is of little use, and correlative study is at poly- also fewer.The material of external structure sound phonetic function assessment
For material generally using monosyllable, these are monosyllabic generally with the form appearance of phoneme pair.Whitehill and Ciocca is to 17 kinds
Phoneme compares influence of the condition to speech articulation and studies, the results showed that gutturophony and non-gutturophony, tone comparison, supply gas with not
It supplies gas, vowel followed by a nasal consonant and non-vowel followed by a nasal consonant, long vowel and short vowel this 6 kinds of phonemes comparison conditions have important shadow to speech articulation
It rings[26]。
The method that the researcher in China studies structure sound locomotion evaluation is also concentrated mainly on subjective evaluation and medical image hand
Above, whether the content of the subjective evaluation mainly morphosis of the structure sound organ of observation patient and motion state be normal for section；It is yellow
Clear ring, Lu Hongyun etc., which are proposed, moves subjective evaluation scheme than more complete structure sound[27]。
In terms of domestic structure sound phonetic function evaluation studies are concentrated mainly on subjective evaluation, and only a small number of researchers
The concept of structure sound and voice is distinguished.Huang Zhaoming etc. proposes " Chinese Articulation ability tests vocabulary " [28], the vocabulary packet
Containing 50 words, speech rehabilitation teacher by the structure sound voices of 50 words of evaluation subject, can thoroughly evaluating subject to 21 initial consonants
With the Articulation ability of 4 kinds of tones, meanwhile, compared by the comparison of 18 phonemes, 37 minimum voices to come the phoneme of assessing subject
Ability.Chen Sanding et al. has carried out the evaluation of standard Chinese initial consonant, simple or compound vowel of a Chinese syllable and tone to 50 deaf youngsters, and it is general to disclose Chinese
The rule of development of deaf youngster's structure sound voice of call, it is further proposed that early, it is sequence, fault-tolerant and consolidate " speech rehabilitation religion
Educate principle[29].Doctor Zhang Jing of East China Normal University has studied the main error trend for listening barrier children in a consonant structure sound, point
The origin cause of formation is analysed, and proposes listen barrier children's consonant phoneme treatment frame accordingly[30]。
2.3, the application study of speech recognition:
The concept of " speech recognition " is born in the fifties in last century, AT&T Labs, the U.S. take the lead in completing it is monosyllabic and
The identification of isolated word is tested[31], core concept is that model is made in the characteristic parameter of extraction voice signal, uses model in identification
Matched method is identified.The continuous development of voice process technology and abundant, has pushed speech recognition in technology significantly
On innovation, the appearance of the feature extraction and matchings algorithm such as linear predictive coding (LPC) and dynamic time warping (DTW), push
The initial stage upsurge of the Research of Speech Recognition.LPC technique efficiently solves the problems, such as the extraction of phonetic feature, and DTW efficiently solves orphan
The problem of speaker's word speed unevenness in vertical word identification, to the speech recognition excellent of speaker dependent.In recent years, speech recognition
It is all taken seriously, and is used widely in external many fields, especially core algorithm reaches its maturity, and many companies are all
It is proposed the SDK used for technical staff's self-developing, the more well-known ViaVoice for having IBM Corporation[32]And Microsoft
Speech SDK[33]Multilingual in the world is supported Deng, these research achievements, has pushed speech recognition technology in global family
Extensive use in front yard, consumer electronics, service industry, education and medical industry.
The country is in terms of to the Research of Speech Recognition of standard Chinese, oneself is achieved achievement abundant, especially technical level
Always immediately following external development.The research project of oneself the multinomial large vocabulary speech recognition of approval of national " 863 " plan, it is therein
" the improvement hidden Markov model of speech recognition " proposes the non-homogeneous hidden Markov model based on duration distribution, is
To an important improvement of the HMM model used in speech recognition.The speech recognition system THEESP energy developed based on this
The discrimination of enough realizations up to 98.7%[34], this is also the domestic optimum level in the current field.Up to the present, voice is known
Other technology is applied to the speech function assessment of disfluency crowd and trains the research of aspect quite deficient at home, and another party
Face, the speech rehabilitation teacher quality and quantity in China is all much unable to satisfy the rehabilitation demands of China disfluency patient at present, speech
Language obstacle crowd is more urgent for the Speech assessment of automation and the demand of rehabilitation equipment.In summary speech recognition skill
The research and development and status of art and structure sound, voice disorder assessment with training be not difficult to find out, the respective theory of two systems, method and
Technological means has been mature on the whole, and is not combined organically but, is reached for the purpose of more crowd's services in need, this
A problem needs emerging cross discipline to go to solve, and how research has the assessment of speech recognition technology and disfluency with rehabilitation
Machine combines, and realizes that computer assisted automation disfluency assessment has social value and realistic meaning with rehabilitation very much.
Leading reference:
[1]Lutsenko S,Barnes NL,Bartee MY,et al.Function and regulation of
human Copper-Transposing ATPases[J].PhysiolRev,2007,87(3):1011-1046.
[2]Ala A,Walker AP,Ashkan K,et al.Wilsons disease[J].Lancet,2007,369
(9559):397-408.
[3]Wilson SAK.Progressive lenticular degeneration:a famklial nervous
disease associated with cirrhosis of the liver[J].Brain,1912,34:295-509.
[4]Yunusova Y,Weismer G,Kent RD,et al.Breath-group intelligibility in
dysarthria:characteristics and underlying correlates[J].J Speech Lang Hear
Res,2005,48(6):12.
[5] Hu Jiyuan, Lv Daping, Wang Gongqiang wait clinical misdiagnosis research [J] Chinese medical of hepatolenticular degeneration miscellaneous
Will, 2001,81 (11): 642-644.
[6] Wang Xiao Yang, research [J] tcm clinical magazine of Bao Yuancheng hepatolenticular degeneration patient's dysarthrosis,
2012,24(3):202-204.
[7]Brabo NC,Cera ML,Barreto SS,et al.Dysarthria in Wilson’s disease:
analysis of two cases in different stages[J].Rev CEFAC[online].2010,12(3):
509-515.
[8]Wang YT,Kent RD,Duffy JR,et al.Dysarthria associated with
traumatic brain injury:speaking rate and emphatic stress[J].J CommunDisord,
2005,38(3):231-260.
[9] evaluation Yu rehabilitation [J] Chinese Scientific Journal of Hearing And Speech Rehabilitation of Li Shengli dysarthrosis, 2009,
32(1):8-12.
[10]Kent RD,Vorperian HK,Kent JF,et al.Voice dysfunction in
dysarthria:application of the Multi-Dimensiona Voice Program[J].J
CommunDisord,2003,36(4):280-306.
[11]Berry WR,Darley FL,Aronson AE.Dysarthria in Wilson’s disease[J].J
Speech Hearing Res,1974,17(2):169-183.
[12] in He Weijia, progress [J] that Li Shengli's dysarthria Speech acoustic level objectively evaluates
State's rehabilitation theory and practice, 2010,16 (2): 118-120.
[13] Yu Xuen, Yang Renmin 132 cases brain MR imaging of hepatolenticular degeneration [J] apoplexy and neurological disease
Magazine, 2007,24 (1): 30-33.
[14] Beijing Li Shengli speech therapy [M]: Huaxia Press, 2003:77-85.
[15]Wilson SAK.Progressive lenticular degeneration:a familial nervous
disease associated with cirrhosis of the liver[J].Brain,1912,34:295-509.
[16] Zhu Yunbo, Song Lei, Tian Zhu wait using dysarthrosis as hepatolenticular degeneration 2 analysis [J] of onset symptoms
Chinese Chinese and western medical science magazine, 2010,8 (3): 28-29.
[17] Liu Wei, Shi Youkun, Li Jing .Wilson 35 dysfunctions of disease assess [J] modern rehabilitation, 2000,4 (9):
1334-1335.
[18]Machado A,Chien HF,Deguti MM,et al.Nenurological manifestations
in Wilson’s disease:Report of 119cases[J].MovDisord,2006,21:2192-2196.
[19] Bi Caiqin, Sun Junqi, Zhao Hong wait rehabilitation nursing [J] China of hepatolenticular degeneration patient's dysarthrosis
Modern Nursing magazine, 2011,17 (22): 2609-2613.
[20] Li Huan dysarthrosis evaluation studies commentary [J] China special education, 2010,6:59-62.
[21]EnderbyPM.Frenchay Dysarthria Assessment[M].San Diego:Colifornia
College-hill Press,1983:34-53.
[22] Clinics and Practices [J] the massage and medical science of recovery therapy of Mao Jianming dysarthria, 2010,36:54-55
[23]Altmann G..Prolegomena to Menzerath’s Law[J].Glottometrika,1980,
2:1-10.
[24] Beijing Wu Zongji, Lin Maocan experimental phonetics summary [M]: Higher Education Publishing House, 1989:153-190.
[25]Mermelstein.P..Articulatory model for the study of speech
production[J].Journal of Acoustic Society of America,1973,61(2):581-587.
[26]Whitehill.T.L,,Ciocca V..Speech errors in Cantonese speaking
adults with cerebral palsy[J].Clinical Linguistics&Phonetics,2000,14:111-130.
[27] the province spy school Lu Hongyun, Huang Zhaoming, Zhou Hong speech rehabilitation specialized instrument and equipment configuration standard [J] the modern times are special
Education, 2010, (6): 31-34.
[28] Shanghai Huang Zhaoming, Du Xiaoxin speech therapy [M]: publishing house, East China Normal University, 2014.
[29] Chen Sanding, Xu Changhong, the rule of the deaf virgin Phonetic Speech Development of Li Yuming Han nationality and Rehabilitation from Illness with Counterplan research [J] China
Rehabilitation, 1996,11 (2): 53-54.
[30] Zhang Lei listens analysis and the therapeutic strategy [master thesis] of barrier children's initial consonant structure sound exception, Shanghai: East China
Normal university, 2009.
[31]K.H.Davis,R.Biddulph,S.Balashek.Automatic Recognition of Spoken
Digits[J].Journal of the Acoustic Society of America,1952,24(6):637-642.
[32]International Bussiness Machine.IBM Desktop ViaVoice.2003,http://
www-01.ibm.com/software/pervasive/viavoice.html
[33]Microsoft.Microsoft Speech Platform-Software DevelopmentKit.2010,
Http:// www.microsoft.com/en-us/download/details.aspx? id=14373.
[34] improvement [J] electronic letters, vol of Zhan Puming, Wang Zuoying, Lu great Jin speech recognition Hidden Markov Model,
1994,22(1):9-15。
Summary of the invention
The present invention devises a kind of self-evaluating system of compressed sensing based hepatolenticular degeneration disfluency, solves
The technical issues of be how to organically combine the assessment of speech recognition technology and disfluency and rehabilitation, artificial intelligence application is arrived
In clinical medicine, realizes computer assisted automation disfluency assessment and rehabilitation, can not only assist diagnosis, control
It treats, it can also be used to which patient voluntarily diagnoses, and for the Finding case state of an illness and the development of the grasp state of an illness in time has great help.
In order to solve above-mentioned technical problem, present invention employs following scheme:
A kind of self-evaluating system of compressed sensing based hepatolenticular degeneration disfluency, including dysarthrosis are commented automatically
Estimate module, disfluency state of an illness categorization module, the automatic categorization module of the state of an illness, state of an illness grade separation module.Dysarthrosis is commented automatically
Estimating functional module includes the judgement read word, read word, tone, and system randomly selects the entry in vocabulary and pronounced by subject, herein
Gender, the levels such as illness type need to be chosen.Specific the barrer types will be provided after system evaluation, and generate test report.Speech barrier
Hindering state of an illness categorization module includes " voice repetition ", " voice switching " evaluation module, and system randomly selects entry and language in vocabulary
Sentence is pronounced by subject, this module can be in conjunction with the evaluation of Frenchay dysarthrosis rating method and clinician, by result
It is recorded in table, record result in detail is also had after system evaluation.The automatic categorization module of the state of an illness, by selecting initial consonant, simple or compound vowel of a Chinese syllable, word
The overall merit of the various aspects such as word, phrase and sentence in table, provides test report, and may recognize that dysarthria
The cause of disease.State of an illness grade separation module carries out grade distinction mainly for hepatolenticular degeneration, by investigating subject to system word
The pronunciation situation of word evaluates state of an illness grade in table.
A kind of self-evaluating system of compressed sensing based hepatolenticular degeneration disfluency, it is characterised in that:
Disfluency patient voice caused by different syndromes is acquired, sub-category speech database is established；
Using word as the formulation of the experiment vocabulary of object；
The structure sound function of patient is carried out in terms of content, tone and the barrer types three by equipment such as computers automatic
Evaluation；
On the basis of Speech SDK identifies engine, in conjunction with the identification engine voluntarily constructed, it is automatic to develop dysarthrosis
Assess feasibility analysis system；
It is put down using the speech recognition algorithm of Hidden Markov Model from whole based on the small feature of sample amount early period
Equal angle realizes optimal identification process；
By the extraction to different sexes, all ages and classes stage voice spectrum signal characteristic parameter, pitch period, frequency are discussed
The difference of rate, short-term spectrum, short-time energy, the short-time average magnitude supplemental characteristic in different syndromes, then by the fundamental frequency of sound sample,
The indexs such as frequency perturbation quotient, Shimmer quotient analyze the difference discussed between various the nervous system disease disfluencys；
Disfluency patient is tested into information and retains storage, after the treatment of a period of time, is detected again in contrast,
Analysing patient's condition development.
A kind of self-evaluating system of compressed sensing based hepatolenticular degeneration disfluency, including
It is proposed the preprocess method of reasonable WD patient's voice signal；
Establish WD patient's voice, image, video, medical history and CT magnetic resonance data database；
Obtain the characteristic attribute of the speech characteristic parameter for WD patient's disfluency, and structure on the basis of existing vocabulary
Build the effective tailored version vocabulary of more specification；
It proposes feasible Classification of Speech recognizer, realizes the division of state of an illness grade；
Research and development can satisfy oneself of the compressed sensing based hepatolenticular degeneration disfluency of clinical practice application requirement
Assessment system.
Further, the preprocess method of WD patient's voice signal, comprising the following steps: step 1: the input of sound with
Filtering；Step 2, the compressed sensing of voice signal；Step 3, end-point detection；Step 4, preemphasis；Step 5, adding window and framing.
Further, filtering adds bandpass filter to voice signal in the step 1, added bandpass filter
Upper and lower cutoff frequency is respectively fH=20kHz, fL=60Hz.
Further, in the step 3 by the autocorrelation maximum of voice signal and cross that threshold rate combines in the way of it is real
The end-point detection of existing efficient voice: assuming that the time-domain expression after voice signal adding window is x (m), wherein n-th frame signal representation
Formula is xn(m), frame length N, the then short-time autocorrelation function of the voice signal are as follows:
Wherein, 0≤k≤K, K are the largest delay points.
Further, step 4 purpose is to increase the energy of high fdrequency component, signal-to-noise ratio is effectively improved, so as to letter
It number carries out that unified signal-to-noise ratio can be used when spectrum analysis and channel parameters calculate, reduces difficulty in computation.Using single order number
Filter composition can improve the preemphasis digital filter of signal high fdrequency component with 6dB/ octave:
H (z)=1- μ z-1 (3)
System function H (z) is z-1Multinomial, its pole on the origin of Z plane, wherein μ be pre emphasis factor.
Further, window function is to be denoted as ω for the tool of the extraction " speech frame " from continuous speech in the step 5
(n)；The characteristic of window function is exactly the voice segments whole zero setting that will be needed except processing region, i.e., to voice signal " framing ".Point
There are two ways to frame: continuously framing and overlapping framing, overlapping framing are the framing methods generallyd use；The voice signal of adding window
Expression formula are as follows:
sω(n)=s (n) g ω (n) (4)
ω (n) is window function, and s (n) is primary speech signal.
Further, the speech characteristic parameter of WD patient's disfluency be short-time energy, Mel cepstrum coefficient (MFCC) and
WD patient's voice nonlinear characteristic parameters.
The self-evaluating system of the compressed sensing based hepatolenticular degeneration disfluency has following
The utility model has the advantages that
(1) present invention proposes the compressed sensing sampling of improved voice signal, has provided to remote patient data sampling
The hardware supported and data compression technique of power are of great immediate significance for the remote collection of realization data with test, are
The acquisition of each area sample data needed for the building of later data library lays the foundation, and also makes for automatic identification system to all patients
With providing powerful technical support.
(2) present invention studies traditional vocabulary of disfluency, by mathematical tool to the speech for influencing WD patient
The voice signal of obstacle carries out step analysis, extracts characteristic attribute, establishes the tailored version vocabulary for being suitable for WD patient.
(3) present invention building WD database proposes that the various aspects sufferer feature of patient, which is carried out classification with data, begs for
By, be applied to automatic identification system in, have important practical significance and accurate data basis.
Detailed description of the invention
Fig. 1: the block diagram of the extraction process of Mel cepstrum coefficient in the present invention.
Fig. 2: speech processes schematic diagram in the present invention.
Fig. 3: experimental comparison figure of the present invention；
Fig. 4: the present invention is based on the self-evaluating working-flow figures of the hepatolenticular degeneration disfluency of compressed sensing.
Specific embodiment
Below with reference to Fig. 1 to Fig. 3, the present invention will be further described:
Evaluation for current dysarthrosis does not still determine quantifier elimination, and what the research of WD dysarthrosis was more a lack of shows
Shape is proposed first to sample voice signal using compressed sensing technology, recycles a variety of points of random forest, SVM classifier etc.
Class knows method for distinguishing and carries out Classification and Identification to the state of an illness of WD patient, and studies the key technology that it is realized.
The invention mainly comprises four big functional modules:
The automatic evaluation module of dysarthrosis, the automatic evaluation module of voice disorder, state of an illness classification automatic identification module, state of an illness etc.
The automatic evaluation module of grade.
The automatic evaluation function module of dysarthrosis includes the judgement read word, read word, tone, and system randomly selects in vocabulary
Entry is pronounced by subject, needs to choose gender, the levels such as illness type herein.Specific the barrer types will be provided after system evaluation,
And generate test report.
The automatic evaluation function module of voice disorder includes " voice repetition ", " voice switching " evaluation module, and system is selected at random
Entry and sentence in vocabulary is taken to be pronounced by subject, this module result can be recorded in table, system in conjunction with subjective judgement
Also record result in detail is had after evaluation and test.
The state of an illness is classified automatic identification module, and the various aspects such as selection initial consonant, simple or compound vowel of a Chinese syllable, word, phrase and sentence in vocabulary are passed through
Overall merit, provide test report, and may recognize that the cause of disease of dysarthria.
The automatic evaluation module of state of an illness grade carries out grade distinction mainly for hepatolenticular degeneration, by investigating subject
State of an illness grade is evaluated to the pronunciation situation of word in system vocabulary.
1.2, target of the invention:
Great demand of the present invention towards WD patient clinical diagnoses and illness analysis, by being quantified to patient's disfluency
Analysing and evaluating is that provide accurate underlying parameter be final goal for WD cerebral injury in patients degree and state of an illness rank, is felt around compression
Know that the key scientific problems in the sampling of voice signal and the classification and identification algorithm of Small Sample Database conduct a research, it is specific to study
Target is as follows:
(1) reasonable Speech Signal Compression cognitive method is proposed.
(2) database of the data such as WD patient's voice, image, video, medical history and CT magnetic resonance is established.
(3) the speech parameter characteristic attribute for being directed to WD patient's disfluency is obtained, and is constructed on the basis of existing vocabulary
The effective tailored version vocabulary of more specification.
(4) it proposes feasible Classification of Speech recognizer, realizes the division of state of an illness grade.
(5) research and development can satisfy the WD disfluency classification automatic identification system of clinical practice application requirement.
1.3.1, the pretreatment of voice signal:
Voice pre-treatment step of the present invention includes: filtering, compression sampling, conversion, preemphasis, adding window, framing
And etc..
(1) input and filtering of sound；
By using the voice of the one-way type microphone samples patient of high-quality, as input signal of the invention.Filtering
It is to inhibit frequency values in input signal is more than the frequency component of sample frequency, to prevent aliasing from interfering；Meanwhile also to press down
The interference of 50Hz alternating current working frequency processed.Since the speech sample frequency of high-quality is usually 44100Hz, needed in voice signal
Inhibiting frequency is more than the frequency component of 22050Hz, therefore, the step for filtering is actually to voice signal addition band logical filter
Wave device, in the present invention, the upper and lower cutoff frequency of added bandpass filter is respectively fH=20kHz, fL=60Hz.
(2) compressed sensing of voice signal；
Due to the time-varying characteristics of voice signal, need to carry out Short Compression and perception to signal.
The analysis of voice signal is usually exactly to carry out linear prediction analysis to the sinusoidal model of voice signal or to signal x
The formant and frequency information of decomposition carry out cepstral analysis.To the rarefaction of voice signal, there are many kinds of research methods: base at present
This redundant dictionary, based on approximate KLT in method, discrete cosine transform (DCT), and the voice signal based on wavelet basis is sparse
Representation method etc..Rarefaction representation shaped like x=ψ α in compressive sensing theory, voice signal can use the dct transform shaped like formula (1)
To replace:
X=c-1·θ1 (1)
C herein is a real transformation matrix, θ1It is DCT coefficient.
In the application of CS compressed sensing, widely using random matrix as sampling base Φ；This is because CS compression sense
Know the incoherence for needing height, and random matrix and any base all have very big incoherence.Bai Gaosi uniformly makes an uproar
Sound all has good sampling base, therefore the random matrix of a usually used independent same distribution (IID) to CS compressed sensing,
Including random gaussian matrix, Teoplitz and circular matrix etc..
The reconstruct of signal has critical role in CS compressive sensing theory, and signal can be from one group of a small amount of measured value
It is reconstructed or restores.By using optimization algorithm, can in the case where receiving end does not need loss information reconstruction signal.CS pressure
The method of contracting sensing reconstructing Resolving probiems is generally divided into two classes, and one is linear program is used, to restore data, (such as base is tracked
), BP another kind is using second order greedy algorithm (such as orthogonal matching pursuit OMP).
(3) end-point detection；
The suitable end-point detecting method of selection can shorten signal processing time, moreover it is possible to it is mixed to exclude invalid section in signal
Noise enhances the recognition performance of system.It generally requires and is combined judgement, this hair using two or more characteristic parameters
It is bright improved to realize efficient voice in such a way that the autocorrelation maximum of voice signal and mistake threshold rate combine using a kind of
End-point detection.Assuming that the time-domain expression after voice signal adding window is x (m), wherein n-th frame signal expression is xn(m),
Frame length is N, then the short-time autocorrelation function of the voice signal are as follows:
Wherein, 0≤k≤K, K are the largest delay points.
(4) preemphasis；
The purpose of preemphasis is the energy of increase high fdrequency component, effectively improves signal-to-noise ratio, to carry out frequency to signal
Unified signal-to-noise ratio can be used when calculating in spectrum analysis and channel parameters, reduces difficulty in computation.Using order digital filter group
At the preemphasis digital filter that can improve signal high fdrequency component with 6dB/ octave:
H (z)=1- μ z-1 (3)
(5) adding window and framing
Window function is the tool for extracting " speech frame " from continuous speech, is denoted as ω (n).The characteristic of window function is exactly
The voice segments whole zero setting except processing region will be needed, i.e., to voice signal " framing ".There are two ways to framing: continuous point
Frame and overlapping framing, overlapping framing are the framing methods generallyd use.The voice signal expression formula of adding window are as follows:
sω(n)=s (n) g ω (n) (4)
Overlapping sub-frame processing is carried out to voice signal using " Hamming window " in the present invention, adding window direction is needed from voice signal
The part of analysis starts, and successively slides along time shaft.
The expression-form for the Hamming window function that one frame length is N are as follows:
1.3.2 the selection of characteristic parameter:
Common identification parameter type includes time domain, frequency domain and transform domain, includes multiple parameters in each type.Wherein
Common feature parameter for identification is as shown in table 1:
1 common feature parameter of table
(1) short-time energy:
Short-time energy is to judge one of most important feature of sound-type, and the present invention chooses logarithmic energy conduct in short-term first
One of characteristic parameter of speech recognition:
(2) Mel cepstrum coefficient (MFCC):
The present invention chooses the another kind of characteristic parameter of Mel cepstrum coefficient (MFCC) as language identification of the present invention.From process
The basic procedure that Mel cepstrum coefficient is extracted in pretreated voice signal is as shown in Figure 1.Firstly, to the language of each windowed function
Sound frame does Fast Fourier Transform (FFT) and obtains power spectrum；Power spectrum is obtained into Mel frequency spectrum, Mel by Mel filter group again
Filter group is actually one group of normalized triangle bandpass filter, i.e., seeks log spectrum to power spectrum；Then, to Mel
Frequency spectrum solves cepstrum, that is, does discrete cosine transform (DCT) and obtain Mel cepstrum, Mel cepstrum coefficient is one group of feature vector, i.e.,
Wherein, N is the number of Mel filter group intermediate cam wave, takes N=20 in the present invention.K=2,3 is taken ..., Ck when 13
12 results as MFCC coefficient.This 12 dimension MFCC coefficient of group is exactly required this frame language for being used for speech recognition of the present invention
The characteristic parameter of sound, as shown in Figure 1.
(3) extraction of WD patient's voice nonlinear characteristic parameters:
The characteristics of for WD patient and voice collecting environment, when selection phonetic feature carrys out characterization information, it is contemplated that in short-term
The phonetic features such as average energy (or amplitude), short-time zero-crossing rate are vulnerable to the distance of patient and microphone, angle and ambient noise
Interference, and durations for speech (word speed) and pure and impure sound changing pattern can be by individual patients language differences (dialect) and subjective pronunciation
Attitude (emotion) is influenced.Therefore, the phonetic feature for being not easy to be affected by other factors should be chosen, at the same cast out other vulnerable to
The feature of interference accordingly reduces its weight.Influence according to fatigue to voice signal, feelings of the same person in hair same voice
Under condition, nonlinear characteristic: maximum Lyapunov exponent (MLE), approximate entropy (ApEn) and fractal dimension and traditional characteristic: base
The variation of voice frequency, formant, sound channel time-varying system parameter can more objectively reflect the voice messaging of patient.
1.4 critical issues intended to solve:
(1) the disfluency test vocabulary for being suitable for WD patient is established.
By the analysis to collected WD patient's speech samples, in Huang Mingzhao-Han Zhijuan vocabulary that is existing and using
On the basis of, the dedicated sample being applicable in suitable for WD patient is designed, in addition, also can for the patient of different geographical, different dialects
Design specific vocabulary.
(2) the speech parameter feature extraction of WD disfluency and database building.
Collected WD patient's speech samples are carried out with the signature analysis of voice signal, extracts the spy that typical case has experimental value
Sign analyzes the difference of the speech characteristic parameter of different patients or different state of an illness grades, clearly influences the feature of state of an illness grade classification
Parameter rationally quantifies the correlativity between each characteristic parameter and state of an illness grade by emulation experiment, and then extracts special
Attribute is levied, while constructing patient's speech database.
(3) research of classification of speech signals recognizer.
Classification and Identification is carried out for WD patient's voice signal data of acquisition, first by the method for machine learning from sample
Wherein valuable rule is found in data, from database analysis treatment as a result, the grade of the state of an illness, from training examples
The algorithm of middle study objective function.By training sample, data are classified using the classifier of algorithms of different, to reach
The state of an illness is classified.
2.1 the quasi- research method taken:
(1) method for taking theoretical research and experimental study to combine.
(2) it is based on compressive sensing theory, using sparse dictionary and convex optimized algorithm is solved, in conjunction with to WD patient's voice signal
Research and analyse, compression sampling is carried out to the voice signal of WD patient.
(3) random forest, neural network, supporting vector are studied by theory analysis and machine learning by experimental study
A variety of classification and identification algorithms such as machine SVM, and the kernel function for optimization of analyzing and researching.
(4) by theory analysis and numerical value analysis means, how research is classified the state of an illness of WD patient, and to vocabulary
It is updated and sets, the overall data library of WD patient is expanded and safeguarded.
2.2, technology path, as shown in Figure 2: including preprocessing module, identification module and output module.Wherein pre-process
Pass through A/D conversion, end-point detection, preemphasis, adding window and parameter extraction after voice input in module, identification module includes standard
Acoustics template and identification are identified that identification is based on dictionary and language after parameter extraction again after the processing of standard acoustic template
Method rule.Recognition result is exported by the output interface of output module.
2.3 experimental programs:
(1) design of speech recognition main performance index
Vocabulary
Selected entry can reflect the mandarin Articulation ability of speaker comprehensively.
Sound combination including 21 initial consonants and 18 simple or compound vowel of a Chinese syllable phonemes.
Preliminary " Chinese Articulation ability assesses vocabulary " (2006) for formulating selection Huang Mingzhao-Han Zhijuan establishment are used as reference.
In conjunction with the different expression form of the nervous system disease disfluency, vocabulary is grouped, it is final to determine vocabulary tool
The composition of 50 words of body.
Acquisition mode: subject per says the entry in a vocabulary, belongs to isolated word recognition.
The identification based on Microsoft's SAPI continuous speech recognition is combined to draw using the speech recognition engine based on isolated word recognition
Hold up common completion.
Identify object --- unspecified person
Noise requirements: the noiseproof feature of system needs to adapt to environmental change, it is also necessary in noiseless or less noise peace and quiet
It is used in environment, needs higher robustness.
(2) design of experimentation:
The more apparent patient of disfluency symptom is as subject in screening the nervous system disease such as disturbances in patients with Parkinson disease.
Before the essential information (name, height, weight, age etc.), sick time, testing time, symptom and the experiment that record patient
Medicining condition.The specific requirement that experiment is lectured from main examiner person to subject, makes it fully understand the concrete form of experiment.Experiment
When beginning, the distance between microphone and subject lip about 10cm are kept.
Experimental facilities is made of the professional microphone of high-quality one-way fashion, voice collecting card, computer, monitoring grade earphone.
When doctor checks, to room and facility requirements: answering peace and quiet in room, may not disperse the object of patient's attention
Product.Two should be placed without armchair and a training station.The height of chair is subject to examiner and patient is in same level.For
Patient's dispersion attention is avoided, the relatives or nursing staff of patient not accompany indoors.
The room that when voice collecting and system are evaluated and tested in real time, the selection of sample plot point is quiet, does not interfere with, ambient noise are less than
45dB。
Specific steps:
Disfluency patient voice caused by different syndromes is acquired, sub-category speech database is established.
Using word as the formulation of the experiment vocabulary of object.According to being instructed for voice disorder for the propositions such as Liu Qiaoyun, Huang Zhao ring
Experienced " CRDS " strategy, formulates the vocabulary for being suitable for this experiment.
The structure sound function of patient is carried out in terms of content, tone and the barrer types three by equipment such as computers automatic
Evaluation.
On the basis of Speech SDK identifies engine, in conjunction with the identification engine voluntarily constructed, it is automatic to develop dysarthrosis
Assess feasibility analysis system.
It is put down using the speech recognition algorithm of Hidden Markov Model from whole based on the small feature of sample amount early period
Equal angle realizes optimal identification process.
By the extraction to different sexes, all ages and classes stage voice spectrum signal characteristic parameter, pitch period, frequency are discussed
The difference of rate, short-term spectrum, short-time energy, the short-time average magnitude supplemental characteristic in different syndromes, then by the fundamental frequency of sound sample,
The indexs such as frequency perturbation quotient, Shimmer quotient analyze the difference discussed between various the nervous system disease disfluencys.
Pass through rhythm structure (word length, morphology and tone), phoneme and phonetic feature (syllable, tone, consonant, vowel, lexeme
Deng) etc. subregions the similarities and differences of different type patient disfluency are discussed, establish the sound bank of region dialect, understand dialect with it is general
The difference of call.
Subject is tested into information and retains storage, after the treatment of a period of time, is detected again in contrast, analysis disease
Feelings development.
As shown in figure 3, being fallen to predictive coefficient, cepstrum, LPC respectively firstly, carry out Experimental comparison to the pronunciation of Chinese character " west "
The parameters such as the Power estimation that spectrum, short-time spectrum, LPC Power estimation, Line Spectral Pair coefficients and line spectrum pair are asked are tested.
As shown in figure 4, a kind of self-evaluating system of compressed sensing based hepatolenticular degeneration disfluency, to different diseases
Disfluency patient voice caused by disease is acquired, and establishes sub-category speech database；
Using word as the formulation of the experiment vocabulary of object；
The structure sound function of patient is carried out in terms of content, tone and the barrer types three by equipment such as computers automatic
Evaluation；
On the basis of Speech SDK identifies engine, in conjunction with the identification engine voluntarily constructed, it is automatic to develop dysarthrosis
Assess feasibility analysis system；
It is put down using the speech recognition algorithm of Hidden Markov Model from whole based on the small feature of sample amount early period
Equal angle realizes optimal identification process；
By the extraction to different sexes, all ages and classes stage voice spectrum signal characteristic parameter, pitch period, frequency are discussed
The difference of rate, short-term spectrum, short-time energy, the short-time average magnitude supplemental characteristic in different syndromes, then by the fundamental frequency of sound sample,
The indexs such as frequency perturbation quotient, Shimmer quotient analyze the difference discussed between various the nervous system disease disfluencys；
Disfluency patient is tested into information and retains storage, after the treatment of a period of time, is detected again in contrast,
Analysing patient's condition development.
Above in conjunction with attached drawing, an exemplary description of the invention, it is clear that realization of the invention is not by aforesaid way
Limitation, as long as use the inventive concept and technical scheme of the present invention carry out various improvement, or it is not improved will be of the invention
Conception and technical scheme directly apply to other occasions, be within the scope of the invention.
Claims (2)
1. a kind of self-evaluating system of compressed sensing based hepatolenticular degeneration disfluency, it is characterised in that:
Including the automatic evaluation module of dysarthrosis, disfluency state of an illness categorization module, the automatic categorization module of the state of an illness, state of an illness ranking score
Generic module；
The automatic evaluation function module of dysarthrosis includes the judgement read word, read word, tone, and system randomly selects the entry in vocabulary
Pronounced by subject, needs to choose gender, the levels such as illness type herein.Specific the barrer types will be provided after system evaluation, and raw
At test report；
Disfluency state of an illness categorization module includes " voice repetition ", " voice switching " evaluation module, and system randomly selects in vocabulary
Entry and sentence pronounced by subject, this module can in conjunction with Frenchay dysarthrosis rating method evaluate, result be recorded
Record result in detail is also had into table, after system evaluation；
The automatic categorization module of the state of an illness is commented by the synthesis of the various aspects such as selection initial consonant, simple or compound vowel of a Chinese syllable, word, phrase and sentence in vocabulary
Valence provides test report, and may recognize that the cause of disease of dysarthria；
State of an illness grade separation module carries out grade distinction mainly for hepatolenticular degeneration, by investigating subject to system word
The pronunciation situation of word evaluates state of an illness grade in table.
2. a kind of self-evaluating system of compressed sensing based hepatolenticular degeneration disfluency, it is characterised in that:
Disfluency patient voice caused by different syndromes is acquired, sub-category speech database is established；
Using word as the formulation of the experiment vocabulary of object；
It is commented automatically in terms of content, tone and the barrer types three by structure sound function of the equipment such as computer to patient
Valence；
On the basis of Speech SDK identifies engine, in conjunction with the identification engine voluntarily constructed, develops dysarthrosis and assess automatically
Feasibility analysis system；
Based on the small feature of sample amount early period, using the speech recognition algorithm of Hidden Markov Model, from ensemble average
Angle realizes optimal identification process；
By the extraction to different sexes, all ages and classes stage voice spectrum signal characteristic parameter, discuss pitch period, frequency,
The difference of short-term spectrum, short-time energy, the short-time average magnitude supplemental characteristic in different syndromes, then pass through fundamental frequency, the frequency of sound sample
The indexs such as rate perturbation quotient, Shimmer quotient analyze the difference discussed between various the nervous system disease disfluencys；
Pass through rhythm structure (word length, morphology and tone), phoneme and phonetic feature (syllable, tone, consonant, vowel, lexeme etc.)
Equal subregions discuss the similarities and differences of different type patient disfluency, establish the sound bank of region dialect, understand dialect and common
The difference of words；
Disfluency patient is tested into information and retains storage, after the treatment of a period of time, is detected again in contrast, analysis
Progression of the disease situation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910494049.XA CN110223688A (en) | 2019-06-08 | 2019-06-08 | A kind of self-evaluating system of compressed sensing based hepatolenticular degeneration disfluency |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910494049.XA CN110223688A (en) | 2019-06-08 | 2019-06-08 | A kind of self-evaluating system of compressed sensing based hepatolenticular degeneration disfluency |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110223688A true CN110223688A (en) | 2019-09-10 |
Family
ID=67816270
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910494049.XA Pending CN110223688A (en) | 2019-06-08 | 2019-06-08 | A kind of self-evaluating system of compressed sensing based hepatolenticular degeneration disfluency |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110223688A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110728995A (en) * | 2019-09-19 | 2020-01-24 | 中国康复研究中心 | Voice and music ability evaluation system applied to aphasia after stroke |
| CN111210838A (en) * | 2019-12-05 | 2020-05-29 | 中国船舶工业综合技术经济研究院 | Method for evaluating speech cognitive ability |
| CN112617755A (en) * | 2020-12-28 | 2021-04-09 | 深圳市艾利特医疗科技有限公司 | Speech dysfunction detection method, device, equipment, storage medium and system |
| CN113393831A (en) * | 2020-03-13 | 2021-09-14 | 开曼商阿普力特医疗股份有限公司 | Speech input operation method based on at least diphones and computer program product |
| CN114515138A (en) * | 2022-01-06 | 2022-05-20 | 福州市星康朗语教育科技有限公司 | Language disorder assessment and correction system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2004008295A2 (en) * | 2002-07-17 | 2004-01-22 | Keough Steven J | System and method for voice characteristic medical analysis |
| CN108537194A (en) * | 2018-04-17 | 2018-09-14 | 谭红春 | A kind of expression recognition method of the hepatolenticular degeneration patient based on deep learning and SVM |
| CN108564042A (en) * | 2018-04-17 | 2018-09-21 | 谭红春 | A kind of facial expression recognition system based on hepatolenticular degeneration patient |
| CN108922561A (en) * | 2018-06-04 | 2018-11-30 | 平安科技（深圳）有限公司 | Speech differentiation method, apparatus, computer equipment and storage medium |
| CN109493886A (en) * | 2018-12-13 | 2019-03-19 | 西安电子科技大学 | Speech-emotion recognition method based on feature selecting and optimization |
-
2019
- 2019-06-08 CN CN201910494049.XA patent/CN110223688A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2004008295A2 (en) * | 2002-07-17 | 2004-01-22 | Keough Steven J | System and method for voice characteristic medical analysis |
| CN108537194A (en) * | 2018-04-17 | 2018-09-14 | 谭红春 | A kind of expression recognition method of the hepatolenticular degeneration patient based on deep learning and SVM |
| CN108564042A (en) * | 2018-04-17 | 2018-09-21 | 谭红春 | A kind of facial expression recognition system based on hepatolenticular degeneration patient |
| CN108922561A (en) * | 2018-06-04 | 2018-11-30 | 平安科技（深圳）有限公司 | Speech differentiation method, apparatus, computer equipment and storage medium |
| CN109493886A (en) * | 2018-12-13 | 2019-03-19 | 西安电子科技大学 | Speech-emotion recognition method based on feature selecting and optimization |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110728995A (en) * | 2019-09-19 | 2020-01-24 | 中国康复研究中心 | Voice and music ability evaluation system applied to aphasia after stroke |
| CN111210838A (en) * | 2019-12-05 | 2020-05-29 | 中国船舶工业综合技术经济研究院 | Method for evaluating speech cognitive ability |
| CN111210838B (en) * | 2019-12-05 | 2023-09-15 | 中国船舶工业综合技术经济研究院 | Evaluation method for speech cognition |
| CN113393831A (en) * | 2020-03-13 | 2021-09-14 | 开曼商阿普力特医疗股份有限公司 | Speech input operation method based on at least diphones and computer program product |
| CN113393831B (en) * | 2020-03-13 | 2023-12-26 | 开曼商阿普力特医疗股份有限公司 | Speech input operation method based on at least diphones and computer readable medium |
| CN112617755A (en) * | 2020-12-28 | 2021-04-09 | 深圳市艾利特医疗科技有限公司 | Speech dysfunction detection method, device, equipment, storage medium and system |
| CN114515138A (en) * | 2022-01-06 | 2022-05-20 | 福州市星康朗语教育科技有限公司 | Language disorder assessment and correction system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mesallam et al. | Development of the arabic voice pathology database and its evaluation by using speech features and machine learning algorithms | |
| Islam et al. | A study of using cough sounds and deep neural networks for the early detection of COVID-19 | |
| CN110223688A (en) | A kind of self-evaluating system of compressed sensing based hepatolenticular degeneration disfluency | |
| Breitenstein et al. | The contribution of speech rate and pitch variation to the perception of vocal emotions in a German and an American sample | |
| Patel et al. | Mapping emotions into acoustic space: The role of voice production | |
| Khan et al. | Classification of speech intelligibility in Parkinson's disease | |
| Panek et al. | Acoustic analysis assessment in speech pathology detection | |
| CN106073706B (en) | A kind of customized information and audio data analysis method and system towards Mini-mental Status Examination | |
| Kadi et al. | Fully automated speaker identification and intelligibility assessment in dysarthria disease using auditory knowledge | |
| CN109727608A (en) | A kind of ill voice appraisal procedure based on Chinese speech | |
| Qin et al. | Automatic assessment of speech impairment in cantonese-speaking people with aphasia | |
| Zhang et al. | Intelligent speech technologies for transcription, disease diagnosis, and medical equipment interactive control in smart hospitals: A review | |
| Zakariah et al. | An analytical study of speech pathology detection based on MFCC and deep neural networks | |
| Khan et al. | Assessing Parkinson's disease severity using speech analysis in non-native speakers | |
| Duville et al. | The Mexican Emotional Speech Database (MESD): elaboration and assessment based on machine learning | |
| Dumpala et al. | Detecting depression with a temporal context of speaker embeddings | |
| CN110211566A (en) | A kind of classification method of compressed sensing based hepatolenticular degeneration disfluency | |
| Zhang et al. | Robust vocal quality feature embeddings for dysphonic voice detection | |
| Padmini et al. | Age-Based Automatic Voice Conversion Using Blood Relation for Voice Impaired. | |
| Milani et al. | A real-time application to detect human voice disorders | |
| Mohamad Jamil et al. | A flexible speech recognition system for cerebral palsy disabled | |
| Yingthawornsuk et al. | Distinguishing depression and suicidal risk in men using GMM based frequency contents of affective vocal tract response | |
| He | Stress and emotion recognition in natural speech in the work and family environments | |
| Keskinpala et al. | Screening for high risk suicidal states using mel-cepstral coefficients and energy in frequency bands | |
| Lai et al. | ResNet for recognition of Qi-deficiency constitution and balanced constitution based on voice |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190910 |
|
| RJ01 | Rejection of invention patent application after publication |