EP0836176A2 - Procédé pour la synthèse d'un cadre d'un signal vocal - Google Patents

Procédé pour la synthèse d'un cadre d'un signal vocal Download PDF

Info

Publication number
EP0836176A2
EP0836176A2 EP97116746A EP97116746A EP0836176A2 EP 0836176 A2 EP0836176 A2 EP 0836176A2 EP 97116746 A EP97116746 A EP 97116746A EP 97116746 A EP97116746 A EP 97116746A EP 0836176 A2 EP0836176 A2 EP 0836176A2
Authority
EP
European Patent Office
Prior art keywords
rpe
pulses
excitation
speech
ideal
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.)
Ceased
Application number
EP97116746A
Other languages
German (de)
English (en)
Other versions
EP0836176A3 (fr
Inventor
Udo Görtz
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nokia Oyj
Original Assignee
Nokia Mobile Phones Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nokia Mobile Phones Ltd filed Critical Nokia Mobile Phones Ltd
Publication of EP0836176A2 publication Critical patent/EP0836176A2/fr
Publication of EP0836176A3 publication Critical patent/EP0836176A3/fr
Ceased legal-status Critical Current

Links

Images

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
    • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
    • G10L19/04—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using predictive techniques
    • G10L19/08—Determination or coding of the excitation function; Determination or coding of the long-term prediction parameters
    • G10L19/10—Determination or coding of the excitation function; Determination or coding of the long-term prediction parameters the excitation function being a multipulse excitation
    • G10L19/113—Regular pulse excitation

Definitions

  • Time domain language coders which will be discussed below, work essentially all according to the same principle: a linear synthesis filter is applied with an excitation signal such that its Output signal to be transmitted speech signal in the sense of The approximate error measure to be determined approximates as well as possible. Often it is Excitation signal from two components. The former should be the harmonic, help to reproduce mostly voiced parts of speech, the latter the noise-like Speech components. The actual sound shaping, the real one Language tract through the pharyngeal-oral-nasal cavity happens to be done through the synthesis filter. The achievable voice quality depends significantly from the excitation of the synthesis filter.
  • So-called residual signal coders such as the RPE-LTP speech coders currently used in digital mobile communications, do not achieve the voice quality required today with comparatively low complexity and bit rates well above 10 kB / s.
  • the starting point of the invention is an "ideal RPE sequence becomes, as in his time by P. Kroon in his thesis "time-domain coding of (near) great quality speech at 16 kb / s ". Delft University of Technology, March 1985, determined. First, therefore, is on the determination of the RPE and the variant used in the RPE-LTP encoder this stimulation type received.
  • the impulse response matrix has the following shape
  • an RPE will have only one in each row of M's nonzero element, where the nth line is the position of the nth Pulse of the RPE indicates. If there are m possibilities, by means of L of zero To form an RPE for different pulses also takes the matrix M m different Shape.
  • the "ideal RPE sequence" is the one according to above calculation the error measure E minimized.
  • the determination of the RPE described above requires the release of a coupled, linear system of equations. As defined by the RPE-LTP encoder was not enough computing power available to the Algorithm in a mass market mobile phone to implement. Therefore, a simplified RPE variant is used. After decorrelation filtering of the speech signal to be transmitted remains a residual signal that theoretically in the frequency range of interest having white spectrum. If all spectral components of same intensity, you can look at the transfer of the entire volume dispense, it is enough to transfer the baseband, which one by subsampling the residual signal after previous low-pass filtering wins. This reduces the number of pulses to be transmitted and thus the transmission rate. Decoder side can by interpolation filtering the untransferred high band will be recovered.
  • the values r (0), r (1), ..., r (N-1) represent the current residual signal, r (- (N-1)), r (- (N-2)), ..., r (-1) are values from the signal history.
  • M is exemplified for the case where the first non-zero pulse is at the first position in the RPE vector and every other pulse is nonzero: ⁇ a 0 , 0, a 1 , 0, a 3 , 0, .. a N-2 , 0 ⁇ .
  • M is constructed as indicated above.
  • R, M t M and H are square matrices of the same dimension. The residual signal matrix R can be assumed to be invertible given sufficient speech activity.
  • the impulse response matrix H is also invertible because it is a triangular matrix whose main diagonal is always nonzero.
  • M t M is never invertible; it contains zero columns and zero rows. If z. B. the second, fourth, sixth, ... pulse in the RPE is zero, then there are in the second, fourth, sixth, ... column and row in M t M only zeros.
  • An FIR filter F (z) of length N with which you can see the residual signal before its sampling would have to filter to get the least possible synthesis error, is by specifying the positioning of the non-zero pulses, the synthesis filter, the target signal and the residual signal are not unique certainly. Become arbitrary after filtering the residual signal m pulses set to zero, m linearly independent are missing to determine the N filter coefficients Equations. The rank of A is only as big as the number of zero more different pulses.
  • the error measure used here is also used.
  • the minimization of the error must lead to the same resulting synthesis error in both methods, because the selected error criterion ensures that there is only a minimum apart from edge extremes.
  • the excitation signals of the two exactly identical synthesis filters must thus match exactly in both cases: the vector z from this section and the vector b from the previous section are therefore the same.
  • the filter F (z) is not recalculated when Target signal and impulse response of the synthesis filter have changed.
  • the Filter coefficients are constant.
  • the magnitude frequency response of this filter has the course of a language spectrum regarded as "typical". in this connection it is a low pass with a smooth transition from the passband to the stopband.
  • the cutoff frequency is around 1300 Hz.
  • the filter F (z) can be regarded as a low-pass filter before the scanner. However, due to the smooth transition from fürlledge- into the Barrier area Alias components. Overall, this approach is a fairly rough approximation.
  • the magnitude frequency response of F (z) varies namely not insignificant.
  • the speech signal can be obtained by linear decorrelation filtering not completely decorrelate.
  • the spectrum is therefore not white, but compared to the original spectrum only flatter and overall of lesser intensity.
  • the assumption, already by knowing the baseband Knowing the entire band is a rough approximation and causes especially for speakers with high voices a not insignificant Errors that appear clearly in the RPE-LTP encoder because only the bottom third of the total volume is transmitted, which is the Sub-sampling by a factor of 3 corresponds.
  • Figure 1 shows the CELP principle as it is typically used.
  • a target signal to be approximated is determined by searching (at least) emulated two codebooks.
  • the adaptive codebook (a2) changes depending from the speech signal while the stochastic codebook (a4) is time-invariant is.
  • the search for the best codevectors is done so that not a common, d. H. simultaneous search in the codebooks takes place, as would be required for optimal selection of codevectors but for reasons of cost first the adaptive codebook (a2) is searched.
  • the best codevector according to the error criterion If the best codevector according to the error criterion is found, one subtracts its contribution to the reconstructed target signal from the target vector (target signal) and receives it by a vector from the stochastic Codebook (a4) still to be reconstructed part of the target signal.
  • the search in the individual codebooks follows the same Principle. In both cases, the quotient becomes the square of the correlation the filtered codevector with the target vector and the energy of the filtered target vector for all codevectors. The one codevector, maximizing this quotient is considered the best codevector which minimizes the error criterion (a5).
  • the upstream error weighting (a6) weights the error according to the characteristic of human hearing. Its position is transmitted to the decoder.
  • the CELP principle is characterized in that to find the best Codevector each candidate vector individually filtered (a3) and with must be compared to the target signal. This process causes despite the sequential search of both codebooks a considerable effort, the one proposed in the first CELP publication Codebook size of 1024 vectors not even on powerful Floating-point signal processors in real time. The focus Working on CELP coders has been (and is) busy Therefore, with the question of how to take advantage of the CELP principle can take over without the disadvantage of the high computational effort have to.
  • the invention is based on the object, a method for speech synthesis to create at the specified bitrate on the Scanning stochastic codebooks can completely do without the Affect voice quality and compared without the transmission rate to increase with the use of stochastic codebooks.
  • the synthesis filter coefficients of a tenth-order filter become often in reflection factors or in Line Spectrum Frequencies (LSFs) converted and (vector-) quantized.
  • the excitation of the synthesis filter consists of the weighted superposition of adaptive excitation and the stochastic stimulus together. Both stimulation components are performed sequentially by a more or less suboptimal Codebook search determines the adaptive excitation, so the by excitation of old excitation values recoverable excitation component, is determined first.
  • the degree of suboptimality when searching The codebooks decide on computational effort and voice quality. The goal is to have as few code vectors as possible within the analysis-by-synthesis loop to investigate, thus limiting the computational effort becomes.
  • the new method according to the invention for the determination described here The stochastic stimulus differs significantly from this approach. There is no pre-selection criterion used and the stochastic stimulus is also not vector quantized. It is not about a scalar quantization in the traditional sense, in which one strives is to quantize the transmitted pulses as accurately as possible.
  • the essentials Quality problem with the RPE-LTP encoder is that the RPE one um the factor three subsampled version of the decorrelated speech signal is. Even an exact quantization of the RPE pulses increases the Quality only insignificant. A reduction of the sub-sampling factor on two increases the quality noticeably, but also requires a considerable higher transmission rate. Because the transfer rate of the encoder but not allowed to rise, this path is eliminated.
  • the RPE-LTP coder uses a rather rough long-term prediction, so that the RPE also has to contribute harmonious speech components.
  • the long-term prediction is much more accurate than in the RPE-LTP coder, so that the residual stochastic excitation is essentially noise-like in character and the proper phasing of the stochastic excitation is much more important than accurate quantization of the amplitudes .
  • ACELPs A lgebraic C or E xited L inear P rediction
  • a codebook search answers the question on which pulse positions pulses must be placed.
  • the answer to this question generally causes a considerable effort, even if the codewords consist only of zeros and ones and the signs have already been determined in advance by suboptimal methods.
  • N 20 samples.
  • the resulting amplitudes of the "ideal RPE” are now considered to find the "surviving pulses". At least half of the RPE amplitudes are relatively small. Only a few amplitudes have a large amplitude. It is sufficient to let the large amplitudes survive, they z. B. equal in terms of amount and then only to transfer their position and sign to the decoder. Three to five of the largest pulses in the whole range suffice for good / very good voice quality.
  • the excitation obtained in this way has the form of a pseudo-MPE ( M ulti P ulse E xcitation).
  • the amplitudes of the surviving pulses are all equal or normalized, z. B. to one, so that the sign also equal to the amplitude is what must be communicated to the coder.
  • To determine the excitation is an exact determination of the amplitudes by solving a coupled system of equations not necessarily required. Leave the corresponding pulse positions and signs also refer to a suboptimal solved system. Here come everyone Possibilities considering amplitudes, positions and signs largely preserved the great pulse. One of these possibilities is to determine the pulses sequentially by first the first pulse, whose contribution to the reconstructed target signal from Target signal p is subtracted, then calculates the second pulse, etc.
  • the described method for obtaining a pseudo-MPE from a "ideal” RPE is a combined "closed-loop” / "open-loop” method.
  • the "ideal” RPE is optimal with respect to the target signal to be approximated (closed-loop), while the quantization of the "ideal” RPE does not match Look at this target signal takes place, but from the positions of the maximum Pulse in the RPE vector depends ("open-loop"). This will reduce the computational effort negligible for quantization. That in speech encoders in this bitrate range otherwise usual very expensive Scanning of stochastic codebooks is omitted.
  • Figure 3 shows the speech coder. After sampling the analog Speech signal in block 0 becomes the digital speech signal of a windowing 2 before the LPC analysis 3 to determine the coefficients of the synthesis filter 11, 12 is performed. Purpose of fenestration is the clipping effects by the finite length of the LPC analysis interval to reduce.
  • the synthesis filter is split into two blocks, with block 11 the decay rate of the filter due to the values in the filter memory and block 12 represents the synthesis filter with zero memory at the beginning of each filtering. The superimposition of both output signals is the output of the synthesis filter. Before their quantization 5, the conversion 4 of the direct filter coefficients into "line spectrum frequencies" takes place (LSF), the more favorable properties in terms of Have quantization as direct filter coefficients.
  • the LSFs will be then quantize 5 and the positions in the corresponding LSF code books are transmitted to the decoder.
  • the windowed digital Speech signal is characterized by a volume value 7, which is proportional is the energy contained in the signal. This value becomes logarithmic quantizes 8 and also transmitted to the decoder.
  • the quantized values of the LSFs and the volume are used, as well as in the decoder.
  • the quantized LSFs converted back into direct filter coefficients 6 and how the volume with the corresponding values of the last analysis interval linear interpolated 9. The above-mentioned calculations are made once per analysis frame, which is 160 samples corresponding to 20 ms long.
  • the following calculations are made eight times per analysis frame, ie every 2.5 ms.
  • the first step is to calculate the current target signal, which should be reproduced. To do this, first subtract the Ausschwinganteil of the synthesis filter 11 due to previous Suggestions from the weighted digital speech signal from block 1. Weight filtering emphasizes areas of importance to the hearing in the voice signal. Now the determination of the adaptive excitation takes place a. It is taken from the adaptive codebook 10 which has a particular one Contains number of past excitation values of the synthesis filter. This Codebook 10 changes its contents after each subframe.
  • the excitation vector a selected from the adaptive codebook, whose filtered and scaled with a gain factor (gain 1) Version in the sense of an arbitrarily chosen error criterion, here square error, smallest distance to the target vector p has. After determining the filtered and scaled adaptive excitation a this is subtracted from the target vector p. It remains through the stochastic Excitation vector c to be minimized residual error. This excitation vector c will not be taken from a codebook, as it is with such Encoders is usually common, but from the target signal p and the Impulse response h of the synthesis filter directly calculated: From the above As explained above, the "ideal" RPE is determined in block 13. The excitation generator 14 determines the positions z. B.
  • gain factor gain 1
  • the speech decoder of Figure 4 is instead of the otherwise existing stochastic codebook an excitation generator 24, the o. e. Receives parameters from the speech coder, ie the position of the first of Zero different pulse of the ideal RPE sequence, the positions of the surviving pulses as well as the sign of the surviving pulses. From these Parameters, the stochastic excitation vector c is formed, the is supplied to the synthesis filter 21 after amplification.
  • the processing steps to be performed by the decoder otherwise correspond essentially those that were already running in the coder, with the exception, however, that for the construction of the filter coefficients and the excitation required codevectors from the various Codebooks based on the position information provided by the coder be taken immediately.
  • there is still one Postprocessing of the synthetic speech signal which is at the output of the LPC synthesis filter 21.
  • the post-processing filter 22 emphasizes the regions important to the listening experience in the voice signal and helps Disturbances caused by the coding itself and possible transmission errors have revealed, at least partially conceal.
  • To final D / A conversion 23 is again an analog voice signal to disposal.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Signal Processing (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Compression, Expansion, Code Conversion, And Decoders (AREA)
EP97116746A 1996-10-09 1997-09-25 Procédé pour la synthèse d'un cadre d'un signal vocal Ceased EP0836176A3 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
DE19641619A DE19641619C1 (de) 1996-10-09 1996-10-09 Verfahren zur Synthese eines Rahmens eines Sprachsignals
DE19641619 1996-10-09

Publications (2)

Publication Number Publication Date
EP0836176A2 true EP0836176A2 (fr) 1998-04-15
EP0836176A3 EP0836176A3 (fr) 1999-01-13

Family

ID=7808273

Family Applications (1)

Application Number Title Priority Date Filing Date
EP97116746A Ceased EP0836176A3 (fr) 1996-10-09 1997-09-25 Procédé pour la synthèse d'un cadre d'un signal vocal

Country Status (3)

Country Link
US (1) US6041298A (fr)
EP (1) EP0836176A3 (fr)
DE (1) DE19641619C1 (fr)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6859775B2 (en) * 2001-03-06 2005-02-22 Ntt Docomo, Inc. Joint optimization of excitation and model parameters in parametric speech coders
JP3582589B2 (ja) * 2001-03-07 2004-10-27 日本電気株式会社 音声符号化装置及び音声復号化装置
US6662154B2 (en) * 2001-12-12 2003-12-09 Motorola, Inc. Method and system for information signal coding using combinatorial and huffman codes
JP2008503786A (ja) * 2004-06-22 2008-02-07 コーニンクレッカ フィリップス エレクトロニクス エヌ ヴィ オーディオ信号の符号化及び復号化
JP5596341B2 (ja) * 2007-03-02 2014-09-24 パナソニック インテレクチュアル プロパティ コーポレーション オブ アメリカ 音声符号化装置および音声符号化方法

Family Cites Families (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO1990013112A1 (fr) * 1989-04-25 1990-11-01 Kabushiki Kaisha Toshiba Codeur vocal
JP3054438B2 (ja) * 1989-05-11 2000-06-19 テレフオンアクチーボラゲツト エル エム エリクソン 線型予測音声符号器における音源パルスの位置決め方法
US5091945A (en) * 1989-09-28 1992-02-25 At&T Bell Laboratories Source dependent channel coding with error protection
CA2010830C (fr) * 1990-02-23 1996-06-25 Jean-Pierre Adoul Regles de codage dynamique permettant un codage efficace des paroles au moyen de codes algebriques
US5701392A (en) * 1990-02-23 1997-12-23 Universite De Sherbrooke Depth-first algebraic-codebook search for fast coding of speech
DE9006717U1 (de) * 1990-06-15 1991-10-10 Philips Patentverwaltung GmbH, 22335 Hamburg Anrufbeantworter für die digitale Aufzeichnung und Wiedergabe von Sprachsignalen
FI98104C (fi) * 1991-05-20 1997-04-10 Nokia Mobile Phones Ltd Menetelmä herätevektorin generoimiseksi ja digitaalinen puhekooderi
FI90477C (fi) * 1992-03-23 1994-02-10 Nokia Mobile Phones Ltd Puhesignaalin laadun parannusmenetelmä lineaarista ennustusta käyttävään koodausjärjestelmään
FI95085C (fi) * 1992-05-11 1995-12-11 Nokia Mobile Phones Ltd Menetelmä puhesignaalin digitaaliseksi koodaamiseksi sekä puhekooderi menetelmän suorittamiseksi
FI98164C (fi) * 1994-01-24 1997-04-25 Nokia Mobile Phones Ltd Puhekooderin parametrien käsittely tietoliikennejärjestelmän vastaanottimessa
FI98163C (fi) * 1994-02-08 1997-04-25 Nokia Mobile Phones Ltd Koodausjärjestelmä parametriseen puheenkoodaukseen
US5602961A (en) * 1994-05-31 1997-02-11 Alaris, Inc. Method and apparatus for speech compression using multi-mode code excited linear predictive coding
FR2729245B1 (fr) * 1995-01-06 1997-04-11 Lamblin Claude Procede de codage de parole a prediction lineaire et excitation par codes algebriques
ATE192259T1 (de) * 1995-11-09 2000-05-15 Nokia Mobile Phones Ltd Verfahren zur synthetisierung eines sprachsignalblocks in einem celp-kodierer

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
M. Delprat et al.: "A 6 KBPS Regular Pulse CELP Coder For Mobile Radio Communications", Proceedings of the Workshop on Speech Coding for Telecommunications, Boston (USA), Seiten 179-188 (1991) *

Also Published As

Publication number Publication date
US6041298A (en) 2000-03-21
DE19641619C1 (de) 1997-06-26
EP0836176A3 (fr) 1999-01-13

Similar Documents

Publication Publication Date Title
DE69227401T2 (de) Verfahren zum Kodieren und Dekodieren von Sprachsignalen
DE19604273C2 (de) Verfahren und Vorrichtung zum Durchführen einer Suche in einem Kodebuch im Hinblick auf das Kodieren eines Klangsignales, Zellkommunikationssystem, Zellnetzwerkelement und mobile Zell-Sender-/Empfänger-Einheit
DE4492048C2 (de) Vektorquantisierungs-Verfahren
DE60006271T2 (de) Celp sprachkodierung mit variabler bitrate mittels phonetischer klassifizierung
DE69331079T2 (de) CELP-Vocoder
DE69125775T2 (de) Sprachkodierungs- und Dekodierungssystem
DE60121405T2 (de) Transkodierer zur Vermeidung einer Kaskadenkodierung von Sprachsignalen
DE69521164T2 (de) System zum Kodieren und Dekodieren von Signalen
DE60225400T2 (de) Verfahren und Vorrichtung zur Verarbeitung eines dekodierten Sprachsignals
DE69604729T2 (de) Verfahren zur sprachkodierung mittels linearer prädiktion und anregung durch algebraische kodes
DE3784942T2 (de) Duplex-datenuebertragung.
DE69726525T2 (de) Verfahren und Vorrichtung zur Vektorquantisierung und zur Sprachkodierung
DE69329569T2 (de) Digitale Kodierung von Sprachsignalen
DE69314389T2 (de) Zweimoden langzeitprädiktion in sprechkodierung
DE69720861T2 (de) Verfahren zur Tonsynthese
DE69033510T3 (de) Numerischer sprachcodierer mit verbesserter langzeitvorhersage durch subabtastauflösung
DE69729527T2 (de) Verfahren und Vorrichtung zur Kodierung von Sprachsignalen
DE69426860T2 (de) Sprachcodierer und Verfahren zum Suchen von Codebüchern
DE60309651T2 (de) Verfahren zur Sprachkodierung mittels verallgemeinerter Analyse durch Synthese und Sprachkodierer zur Durchführung dieses Verfahrens
DE69615870T2 (de) Sprachkodierer mit aus aktuellen und vorhergehenden Rahmen extrahierten Merkmalen
DE4491015C2 (de) Verfahren zum Erzeugen eines Spektralrauschbewertungsfilters zur Verwendung in einem Sprachcoder
DE69702261T2 (de) Sprachkodierung
DE69612788T2 (de) Kodebuch-Suchverfahren für die Sprachverarbeitung
DE69324732T2 (de) Selektive Anwendung von Sprachkodierungstechniken
DE60016305T2 (de) Verfahren zum Betrieb eines Sprachkodierers

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

AK Designated contracting states

Kind code of ref document: A2

Designated state(s): DE FR GB SE

PUAL Search report despatched

Free format text: ORIGINAL CODE: 0009013

RHK1 Main classification (correction)

Ipc: G10L 5/06

AK Designated contracting states

Kind code of ref document: A3

Designated state(s): AT BE CH DE DK ES FI FR GB GR IE IT LI LU MC NL PT SE

17P Request for examination filed

Effective date: 19981211

AKX Designation fees paid

Free format text: DE FR GB SE

RIC1 Information provided on ipc code assigned before grant

Free format text: 7G 10L 19/10 A

GRAG Despatch of communication of intention to grant

Free format text: ORIGINAL CODE: EPIDOS AGRA

17Q First examination report despatched

Effective date: 20020115

RAP1 Party data changed (applicant data changed or rights of an application transferred)

Owner name: NOKIA CORPORATION

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE APPLICATION HAS BEEN REFUSED

18R Application refused

Effective date: 20020705