1 / 61

Dr. Dóra Maros Institute of Telecommunication Óbuda University Telecommunications Techniques II.

Dr. Dóra Maros Institute of Telecommunication Óbuda University Telecommunications Techniques II. Source Coding ( speech ) Error detection and Error correction. Transmission from source to destination. Source ( analog ). Destination ( analog / digital ). Source coding. Source

george
Télécharger la présentation

Dr. Dóra Maros Institute of Telecommunication Óbuda University Telecommunications Techniques II.

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Dr. Dóra Maros Institute of Telecommunication Óbuda University TelecommunicationsTechniques II. SourceCoding (speech) Errordetection and Errorcorrection

  2. Transmissionfromsourcetodestination Source (analog) Destination (analog/digital) Source coding Source decoding -Microphone -Camera -Pressure, temperatureetcsensors -Speaker (ears) -Displays (eyes) -Alarms (see and hear!) Channelcoding Channeldecoding Channel (digital)

  3. Bandwidthallocation Filter A/D Compression Channelcoding Modulation Channel Source information Filter Channel coding Sourceinformation A/D Compression

  4. Howwe „generate” voice? Glottis Vocalcords Vocaltract (resonator filter system) Excitation

  5. Characterization of Human Speech • Frequency: 300-3400 Hz • The charcaterization of a 20 msdurationsampledoesnotchange (excitation, filter system) • Wecanfindsomecorrelationsbetweentwoconsecutive 20 mssamples • Using a longtimeanalysiswecanpredictthenextpattern (differencialcoding, eg. DPCM) • Ourbraincancorrect a lot of „errors”!

  6. Classification of SpeechEncoders Waveformencoder: describestheanalogsignalsinthetime and frequencydomain Vocoders: describestherules of phonation (howwegeneratethevoice) Hybridencoders: combinesgoodproperties of twocoders (lowrate, goodquality)

  7. Subjective MOS MOS: MeanOpinionScore

  8. Theory of WaveformEncoders Sampling (samplingfrequency) Quantization (number of levels) Coding (length of codeword) Benefits: goodvoicequality Disadvantage: bigrate (PCM: 64 kbit/s)

  9. Theory of Vocoders A, E, O, I etc. vowels Excitation B, F, P, M etc. consonants Benefit: lowrate (pár kb/s) Disadvantage: badquality

  10. Voiced - Vowels Typicalamplitude/timesignal Typical PowerSpectrumDensity/ Frequencysignal Pitchfrequency: 50-500 Hz

  11. Voiceless - Consonants Typical PowerSpectrumDensity/ Frequencysignal Typicalamplitude/timesignal (whitenoise)

  12. Sub-bandCoding • Baseband is dividedintosub-bandsregarding PSD • Eachsub-bandsareencodedseparately • Benefits: lownoisesensitivity (differentsamplingfrequencies and codewordlengthareused) • Bitrate: 16-32 kbit/s • Receiver: sub-bandsaresuperimposed (added)

  13. SpeechCodecs MOS AMR: AdaptiveMultirateCodec (3G és 4G mobile) G723: VoIP (mobile) G728: LD-CELP, Low-Delay code excited linear prediction G729: CS-ACELP, Coding of Speech,Adaptive Code ExcitedLinearPrediction (VoIP) GSM FR and EFR: Fullrate, EnhancedFullRate Internet LowBitrateCodec (iLBC)

  14. Theory of HybridEncoder/Decoder AnalysisbySynthesis (AbS)

  15. Excitation: MPE and RPE MPE: MultiplePulseExcitation RPE: RegularPulseExcitation, GSM-13kbit/s RPE MPE MPE: aplitudes and phasesareencoded RPE: onlyamplitudes is encoded

  16. Excitation: CELP (CodeExcitedLinearPrediction) Excitation Table Index Excitation Table • Codetable: 1024 excitationsamples • Onlythe index of table is sent! • Bitrate: 4,8-16 kbit/s

  17. CodecRate tin Encoder Codecrate: tin-tout Typical: 50-100 ms Fast: G728 CELP (Do-CELP) codec, 2-5 ms tout Decoder

  18. GSM SpeechCodec Voiceactivitydetector Waveformencoder Bitrate: 260 bit/20ms, 13kbit/s

  19. VoiceActivityDetector (VAD)

  20. DiscontinousTransmission (DTX) and SID Frame SID: SilenceIndicator

  21. Fletcher-MunsonCurves (sensitivity of ears) Equal-loudness contours for the human ear Phon: Unit of relativeloudness level for pure tones Frequency

  22. ColorSensitivity of Eyes Luminositysensitivity is 3 x more fasterthencolorsensitivity

  23. Compression of SourceInformation Losslesscompression: small part of information is lost Lossycompression:huge part of information is lost CompressionRate: originalinformation/compressedinformation image: lossless : .raw lossy: .jpeg (eg. 25/1)

  24. ErrorCorrection (general) Noisychannel Demodulator Received bit sequence (BER) Errordetectionprocedures Errordetection AutomaticRepeatRequest (ARQ) Errorcorrection ForwardErrorCoding (FEC)

  25. ErrorDetection I. (repetition) • Weusesmallblocks • Eachblock is sentrepeatedly • Efficient, whenwehaveonlyone bit error Sender Receiver 1011 1011 1011 1010 1011 1011 1011 1011 1011 1010 1011 1010

  26. ErrorDetection II. (parity) • Evenoroddparity • Even: we add „1” parityifthenumber of „1”s is odd, otherwiseparityis „0”. • Odd: we add „1” parity, ifthenumber of „1”s is even • Applications: mikrocontollers, databus, datastorage Sender Receiver Evenparity 10110010 10110110 Oddparity 00110010 00110110

  27. ErrorDetection III. (polaritychange) • Wesendtheoriginal and negated bit sequency • Ifthe bit polarity is thesameatthereceiver-error • Disadvantege: bigredundancy • Application: multicarriersystems Sender Receiver Receiver 10110010 10110010 10100010 01001101 01011101 01011101

  28. ErrorDetection IV. (CyclicRedundancyCheck -CRC) • A polimon is generatedfrom bit sequence (data) • Generator polinom is thesame (known) atsender and receiver • CRC= data polinom/generatorpolinom • Application: mobile, RFID, Bluetooth, Ethetnet..etc Sender Receiver MSB LSB 10110010101001 CRC 10100010101001 CRC polinom polinom ? CRC division division generator polinom generator polinom

  29. ErrorDetection V. (CheckSum) • The message is segmentedintoparts • Webinary add theparts (CS) • The message and CS aresent • Webinary add theparts again atreceiverside • IftwoCSsarenotequal-error

  30. ErrorDetection VI. (Hamming coding) Sender Receiver Hamming distance calculator Code i Code i The Hamming distance of thecodewords is greaterorequal d+1 Eg: 4/7 Hamming code …….4 databits – 3 paritybits d=2 errorcan be detected, butonlyone is corrected Number of paritybits = largestexponent (23)

  31. Hamming coding (evenparity) p1: (3,5,7) – 20=1 P2: (3,6,7) – 21=2 P3: (5,6,7) – 22=4 DATA PARITY 7653 p1p2p3 0001 11 0 0010 10 1 0011 011 0100 011 HD:4 7 6 5 4 3 21 HD:3 HD:3

  32. ErrorCorrection (ARQ:AutomaticRepeatRequest) • Fourtypes: • Stop and Wait • Go-Back-N • Selective • Hybrid

  33. Stop and Wait ARQ t1 Data block Sender Receiver t1 Data block Sender Receiver Data block t2 Aftertime out thesenderrepeatsthetransmission Problem: receiverdoesnotknowwhetheritreceivedtherepeatedornextdatablock Solution: 0 and 1 flagbitsinconsecutivedatablocks

  34. Go-Back-N ARQ 1. Wedefine a slidingtimewindow 2. The packetsaresentonebyone (withoutACKs) inthiswindow 3. Ifthefirst ACK is detectedwindowslides. 4. Ifnot, allpacketsaresent again

  35. Selective ARQ Message Long message – one ACK Advantage: if no errors – fasttransmission Disadvantage: Iferrors – themessage is sent again – slowtransmission Smallblock – more ACKs Advantage: werepeateonlythoseblockswhereerrorsaredetected Disadvantage: slow ( ACK)

  36. Hibrid ARQ (Chasecombining) NACK is sent back, receiverdoesnotdropfaultypackets Using an errorscorrectionprocedurewecanrepairthedatafromfaultypackets

  37. Hibrid ARQ (Incrementalredundancy) A FEC (ForwardErrorCoding) is usedatsenderside The redundancybitsaresentafter NACK

  38. ForwardErrorCoding: FEC • Two main types: • Convolutionalcoding • Blockcoding • We add redundancybitswhichhelpcorrectdataatthereceiver

  39. ConvolutionalCoding (theoryexample) • (n,k,m) coder • D laches and modulo2 adders • n: number of outputs • k: number of inputs • m: number of D laches • number of input state : (m-1)2 (3,1,3) coder Generatorpolinoms G1(1,1,1) G2 (0,1,1) G3(1,0,1)

  40. Types of ConvolutionalCoders (Mothercoder and „punctured” coder) Mothercoder:1/n „Punctured” coder: k/n eg: 2/3 coder

  41. Systematiccoder • Wesendoriginaldata (V3,V2,V1 and redundancy (v4)

  42. Non systematiccoder • Wedonotsendoriginaldata • Wesendonlyconvolutionalcodedinfo

  43. Wecode „1” (findtheerror!) Output: 11111011

  44. 1011 vs. 1 111 0 1 11 0 1 0 1 11 ?

  45. Butweuseconvolution! Input bitsRespones (codeddata) 1 11111 0 1 1 0 00000000 (twobits shift) 1 11111 0 1 1 1 11111 0 1 1 ----------------------------------------------------------------------------- Kódolt info 1 111 0 1 11 0 1 0 1 11 Weuse modulo2 adderforoutputs! Shift: L-1 (L: number of inner D laches)

  46. Operationtable

  47. Stategraph Info: 1011 Codedinfo: 1 111 0 1 11 0 1 0 1 11

  48. Treestructure Info: 1011 Codedinfo: 1 111 0 1 11 0 1 0 1 11

  49. TrellisGraph

  50. CodingwithTrellisGraph Info: 1011 Codedinfo: 1 111 0 1 11 0 1 0 1 11

More Related