ATSC Standard: Audio Watermark Emission (A/334)

Transcription

1 ATSC Standard: Audo Watermark Emsson (A/334) Doc. A/334: September 2016 Advanced Televson Systems Commttee 1776 K Stree N.W. Washngton, D.C

2 The Advanced Televson Systems Commttee, Inc., s an nternatonal, non-proft organzaton developng voluntary standards for dgtal televson. The ATSC member organzatons represent the broadcas broadcast equpmen moton pcture, consumer electroncs, computer, cable, satellte, and semconductor ndustres. Specfcally, ATSC s workng to coordnate televson standards among dfferent communcatons meda focusng on dgtal televson, nteractve systems, and broadband multmeda communcatons. ATSC s also developng dgtal televson mplementaton strateges and presentng educatonal semnars on the ATSC standards. ATSC was formed n 1982 by the member organzatons of the Jont Commttee on InterSocety Coordnaton (JCIC): the Electronc Industres Assocaton (EIA), the Insttute of Electrcal and Electronc Engneers (IEEE), the Natonal Assocaton of Broadcasters (NAB), the Natonal Cable Telecommuncatons Assocaton (NCTA), and the Socety of Moton Pcture and Televson Engneers (SMPTE). Currently, there are approxmately 120 members representng the broadcas broadcast equpmen moton pcture, consumer electroncs, computer, cable, satellte, and semconductor ndustres. ATSC Dgtal TV Standards nclude dgtal hgh defnton televson (HDTV), standard defnton televson (SDTV), data broadcastng, multchannel surround-sound audo, and satellte drect-to-home broadcastng. Note: The user's attenton s called to the possblty that complance wth ths standard may requre use of an nventon covered by patent rghts. By publcaton of ths standard, no poston s taken wth respect to the valdty of ths clam or of any patent rghts n connecton therewth. One or more patent holders have, however, fled a statement regardng the terms on whch such patent holder(s) may be wllng to grant a lcense under these rghts to ndvduals or enttes desrng to obtan such a lcense. Detals may be obtaned from the ATSC Secretary and the patent holder. Revson Hstory Verson Date Canddate Standard approved 27 November 2015 Standard approved 19 September 2016 Reference [2] updated to pont to publshed verson of A/342 Part 1: February 2017 Reference [3] updated to pont to publshed verson of A/336: February 2017

3 Table of Contents 1. SCOPE Introducton and Background Organzaton 1 2. REFERENCES Normatve References Informatve References 1 3. DEFINITION OF TERMS Complance Notaton Treatment of Syntactc Structures Reserved Felds Acronyms and Abbrevaton Terms 2 4. SYSTEM OVERVIEW System Archtecture 3 5. SPECIFICATION Physcal Layer Marked Audo Sgnal Marked Audo Presentaton Data Lnk Layer Cell 6

4 Index of Fgures and Tables Fgure 4.1 Generc audo watermark system archtecture Table 5.1 Mappng of Autocorrelaton Dfference Value to Symbol and Bt Values 5 Table 5.2 Parameter Specfcatons 5 Table 5.3 Syntax cell() Structure 6 Table 5.4 Header Sequence 6 v

5 1. SCOPE ATSC Standard: Audo Watermark Emsson 1.1 Introducton and Background Ths document specfes the VP1 audo watermark for use wth systems conformng to the ATSC 3.0 famly of specfcatons. Ths document specfes the format n whch the audo watermark resdes n a PCM audo sgnal. Emsson by a broadcaster of the audo watermark s optonal. A specfcaton for the use of VP1 audo watermarks n emssons wth applcaton to the recovery of ATSC 3.0 servce sgnalng by redstrbuton recevers s provded n A/336 [3]. 1.2 Organzaton Ths document s organzed as follows: Secton 1 Outlnes the scope of ths document and provdes a general ntroducton. Secton 2 Lsts references and applcable documents. Secton 3 Provdes a defnton of terms, acronyms, and abbrevatons for ths document. Secton 4 System overvew Secton 5 System specfcatons 2. REFERENCES All referenced documents are subject to revson. Users of ths Standard are cautoned that newer edtons mght or mght not be compatble. 2.1 Normatve References The followng documents, n whole or n par as referenced n ths documen contan specfc provsons that are to be followed strctly n order to mplement a provson of ths Standard. [1] IEEE: Use of the Internatonal Systems of Unts (SI): The Modern Metrc System, Doc. SI 10, Insttute of Electrcal and Electroncs Engneers, New York, N.Y. [2] ATSC: ATSC Standard: Audo Common Elements, Doc. A/342 Part 1:2017, Advanced Televson Systems Commttee, Washngton, D.C., 24 January Informatve References The followng documents contan nformaton that may be helpful n applyng ths Standard. [3] ATSC: ATSC Standard: Content Recovery n Redstrbuton Scenaros, Doc. A/336:2017, Advanced Televson Systems Commttee, Washngton, D.C., 24 February DEFINITION OF TERMS Wth respect to defnton of terms, abbrevatons, and unts, the practce of the Insttute of Electrcal and Electroncs Engneers (IEEE) as outlned n the Insttute s publshed standards [1] shall be used. Where an abbrevaton s not covered by IEEE practce or ndustry practce dffers from IEEE practce, the abbrevaton n queston wll be descrbed n Secton 3.3 of ths document. 3.1 Complance Notaton Ths secton defnes complance terms for use by ths document: 1

6 shall Ths word ndcates specfc provsons that are to be followed strctly (no devaton s permtted). shall not Ths phrase ndcates specfc provsons that are absolutely prohbted. should Ths word ndcates that a certan course of acton s preferred but not necessarly requred. should not Ths phrase means a certan possblty or course of acton s undesrable but not prohbted. 3.2 Treatment of Syntactc Structures Ths document contans symbolc references to syntactc structures used n the audo, vdeo, and transport codng subsystems. These references are typographcally dstngushed by the use of a dfferent font (e.g., restrcted), may contan the underscore character (e.g., sequence_end_code) and may consst of character strngs that are not Englsh words (e.g., dynrng) Reserved Felds One or more reserved bts, symbols, felds, or ranges of values may be present n ths document. These are used prmarly to enable addng new values to a syntax structure wthout alterng ts syntax or causng a problem wth backwards compatblty, but they also can be used for other reasons. The ATSC default value for reserved bts s 1. There s no default value for other reserved felds. Use of reserved felds except as defned n ATSC Standards or by an ndustry standards settng body s not permtted. See ndvdual feld semantcs for mandatory settngs and any addtonal use constrants. As currently-reserved felds may be assgned values and meanngs n future versons of ths Standard, recevng devces bult to ths verson are expected to gnore all values appearng n currently-reserved felds to avod possble future falure to functon as ntended. 3.3 Acronyms and Abbrevaton The followng acronyms and abbrevatons are used wthn ths document. ATSC Advanced Televson Systems Commttee LSB Least sgnfcant bt. MSB Most sgnfcant bt. bslbf Bt strng, left bt frst. 3.4 Terms The followng terms are used wthn ths document. audo presentaton Has the meanng gven n the ATSC A/342 Audo, Part 1: Common Elements [2]. Also referred to as a Preselecton(DASH-IF), a Presentaton(AC-4), or a Preset(MPEG-H). audo sgnal Has the meanng gven n the ATSC A/342 Audo, Part 1: Common Elements [2]. audo watermark Data whch s embedded n audo essence n such a way that t can be extracted (.e., read) by an approprately desgned extractor. embed The process whereby an audo sgnal s modfed to nclude an audo watermark. embedder A tool or process that s able to embed an audo watermark n an audo sgnal. extractor A tool or process that s able to extract audo watermark packets from an audo sgnal. 2

7 cell A complete transmsson of an ndependently recoverable packet of data n an audo watermark. header A sequence of bts that s present at the start of a cell. marked audo Audo that has an audo watermark embedded n t. PCM Lnear Pulse Code Modulaton. In ths document ths s understood to be the uncompressed format for audo sgnals. symbol The representaton of a bt of bnary nformaton n the audo watermark. VP1 The audo watermarkng technology standardzed n ths specfcaton 4. SYSTEM OVERVIEW 4.1 System Archtecture Fgure 4.1, below, s a block dagram of the basc archtecture for use of VP1 audo watermark technology. At the left of the fgure, pror to dstrbuton of content to recevers, an audo presentaton s nput to a VP1 audo watermark embedder, along wth a sequence of data packets ntended for delvery to a recever va the VP1 audo watermark. The output of the embedder s a marked audo presentaton, whch s ncluded n dstrbuton content. Dstrbuton content s delvered to recevers whch ncorporate a VP1 audo watermark extractor. When a marked audo presentaton s present n the receved audo, the extractor recovers the embedded data packets from the VP1 audo watermark. The VP1 audo watermark s co-resdent wth audo energy n a regon of the audo frequency spectrum contanng perceptually mportant components and s therefore retaned through audo dstrbuton paths. It employs a modulaton and encodng scheme whch s reslent to most types of audo processng, ncludng modern technques for lossy sgnal compresson. It also ncludes error detecton and correcton capabltes. Together, these features permt VP1 to provde a relable auxlary channel for delvery of a data payload to accompany audo and audovsual content through envronments employng heterogeneous audovsual formats, protocols, and nterfaces. Audo watermark embeddng s expected to be performed n broadcast envronments on PCM audo pror to emsson encodng or at an upstream transcode pont. For audo delvered to recevers n an encoded forma audo decodng to PCM audo s expected to be performed pror to audo watermark extracton. Audo Presentaton Watermark Embedder Marked Audo Presentaton Dstrbuton Content Receved Audo Data Packets Watermark Extractor Data Packets Producton / Broadcast Dstrbuton Recever Fgure 4.1 Generc audo watermark system archtecture. 3

8 The VP1 audo watermark s specfed ndependently of the format n whch the audo presentaton s produced; e.g. channel-based, object-based, ambsonc, etc. For any forma the audo presentaton s understood to be comprsed of one or more monophonc audo sgnals that are ntended for synchronzed, smultaneous renderng. The audo watermark s specfed to be present n all synchronous audo sgnals, wth the same symbol embedded at the same renderng tme of each audo sgnal. Wth ths approach, embedded audo watermark data can be extracted from any ndvdual audo sgnal from the audo presentaton, from any subset of audo sgnals from the audo presentaton, or from any lnear combnaton of audo sgnals from the audo presentaton. Ths approach ensures that the audo watermark can be recovered even when multple audo sgnals are mxed together durng decodng, renderng, and/or transcodng of the marked audo presentaton pror to audo watermark extracton. The VP1 audo watermark s specfed n Secton 5 as a sequence of contguous symbols that comprse a contguous sequence of data cells, each contanng an ndvdual data packe embedded across a contnuous tme nterval of an audo presentaton. However, the present document does not set forth any requrement regardng the duraton of the tme nterval that s embedded nor the algnment of ts startng or endng boundares wth a logcal boundary of the audo presentaton or of the broadcast servce (e.g. show segmen advertsemen etc.). 5. SPECIFICATION 5.1 Physcal Layer The physcal layer propertes of the audo watermark are descrbed usng contnuous-tme nomenclature, whch has an unambguous correspondence to dscrete-tme sampled system nomenclature gven the audo samplng rate Marked Audo Sgnal We defne a marked audo sgnal as ss (tt) where t s the contnuous tme varable and s the audo sgnal ndex. The audo watermark shall be present n the markng frequency band of the audo sgnal whch s specfed below n Table 5.2. Let the subband of the marked audo sgnal ss (tt) whch les n the markng frequency band be denoted ss (tt). We defne the autocorrelaton dfference functon, R d (, τ ) as d t+ T / 2 t t+ T R (, τ ) = s '( u) s '( u τ ) du s '( u) s '( u τ ) du t+ T / 2 where t s the tme nstant at whch the calculaton s performed, τ s the autocorrelaton delay, and T s the symbol nterval. Two audo watermark sgnalng modes are specfed, standard sgnalng and nverse sgnalng. For each sgnalng mode, the value of the autocorrelaton dfference functon at tme tt nn = nnnn shall determne the symbol value embedded n the nth symbol nterval as shown below n Table 5.1. Unless otherwse specfed, standard sgnalng mode shall be employed n marked audo. 4

9 Table 5.1 Mappng of Autocorrelaton Dfference Value to Symbol and Bt Values Value of R d (, τ ) Standard Sgnalng Inverse Sgnalng Symbol Value Bt Value Symbol Value Bt Value Greater than or equal to zero Less than zero The parameters of (, τ ) shall be as specfed n Table 5.2 below. R d Table 5.2 Parameter Specfcatons Parameter Descrpton Autocorrelaton Delay, τ Symbol nterval, T Markng Frequency Band Parameter Value 3 mllseconds 1/106 seconds 2.5 khz - 5 khz The magntude of R d (, τ ) employed for a symbol dctates the content-dependent tradeoff between mperceptblty of the audo watermark to the audence and the robustness of the audo watermark to dstorton ntroduced by subsequent audo processng. Ths value s not normatvely specfed. As nformatve gudance to mplementers, acceptable results for typcal use cases of the VP1 audo watermark have been demonstrated by an mplementaton of the technology that employed an average recommended strength for symbols of: avg σσ,nn = 0.3, nn where σ,n denotes the strength of symbol n n audo sgnal and s defned as: 2RR dd (, tt nn, ττ) σσ,nn = EE ss (, tt nn ττ) + EE ss (, tt nn ). where E s (, t) s the energy n the markng frequency band over the symbol nterval startng at tme calculated accordng to: E (, t) = s t+ T t s '( u) Symbol errors may be present n marked audo sgnals conformng wth ths specfcaton as necessary to acheve a desred level of mperceptblty n marked audo presentatons. Marked audo sgnals shall be present only n a marked audo presentaton, as specfed n Secton du 5

10 5.1.2 Marked Audo Presentaton All audo sgnals of a marked audo presentaton shall be synchronously embedded such that the R d (, τ ) for each of the audo sgnals ndcates the same symbol n the same sgnalng mode (.e., standard or nverse) for the same t = tn. For proper detecton, all audo sgnals that are ntended to be decoded and rendered smultaneously must contan dentcal watermarks. If any of the audo sgnals n the audo presentaton are marked, then all of the audo sgnals of that audo presentaton shall contan dentcal watermarks. If there are shared audo sgnals between two audo presentatons, and any of the audo sgnals of the two audo presentatons are marked, then all of the audo sgnals of both audo presentatons shall contan dentcal watermarks. 5.2 Data Lnk Layer Cell A specfc segment of the seres of contguous bts derved from the seres of symbols n the audo sgnal at ncreasng values of t = tn shall form a cell of 159 bts whch are ordered left to rgh correspondng to the tme-orderng of the symbols (frst to last) across a 1.5 second nterval of the audo sgnal. The cell() structure shall have the syntax shown below n Table 5.3. It s dvded nto two regons: header and packet. Table 5.3 Syntax cell() Structure Syntax No. of Bts Format cell() { header 32 bslbf packet 127 bslbf } header A 32-bt header sequence as specfed n Secton , below. packet A 127-bt sequence conveyng a packet of data Header The frst 32 bts of a cell s the header sequence, whch shall have the value as shown below n Table 5.4. Header Table 5.4 Header Sequence Bnary Packet The last 127 bts of a cell s the packet sequence, whch shall convey a 127-bt packet of data. A specfcaton of the contents of a packet wth applcaton of the VP1 audo watermark to the recovery of ATSC 3.0 servce sgnalng by redstrbuton recevers s provded n A/336 [3]. End of Document 6