Data Embedded

Location:

Japan

Posted:

February 19, 2013

Contact this candidate

Resume:

ADPCM

On Data Embedding Methods Applicable to ADPCM Family of Codecs

Kazuhiro KONDO a), Hideki KANAZAWA, Ami YOSHII, Hiroshi KATO,

Tomoyuki SEINO, and Kiyoshi NAKAGAWA

ADPCM G.726 727 7kHz

G.722 ADPCM

ADPCM

ADPCM 7kHz G.726 G.722

Faculty of Engineering, Yamagata University, 4-3-16 Jonan,

Yonezawa, Yamagata 992-8510, Japan

[5]

a) E-mail: ******@**.********-*.**.**

D II Vol. Jxx D II No. xx pp. 1 9 xxxx xx 1

xxxx/xx Vol. Jxx D II No. xx

1 ADPCM

Fig. 1 Con guration of the ADPCM Codec (Trans-

mitter)

3 ADPCM

Fig. 3 Con guration of the Text-Embedded ADPCM

2 ADPCM

Codec by Iwakiri et. al.(Transmitter)

Fig. 2 Con guration of the ADPCM Codec (Re-

ceiver)

ADPCM

ADPCM ADPCM

7kHz

4 ADPCM

ADPCM [6]

[7] Fig. 4 Con guration of the Text-Embedded ADPCM

Codec by Iwakiri et. al.(Receiver)

2 ADPCM

3 7kHz ADPCM i(t)

ADPCM

se(t)

2. 2 ADPCM

2. ADPCM

G.726 ADPCM

2. 1 ADPCM se(t)

1 2 G.726 [2] ADPCM

ADPCM

se(t) s(t) 1 [4]

d(t) = s(t) se(t) i(t) 3 4

i(t) G.726

dq (t) se(t) se(t)

s(t) = dq (t) + se(t) G.726

dq (t n)

ADPCM i(t m)

ADPCM i(t)

ADPCM

6 ADPCM

Fig. 6 Con guration of the Proposed Text-Embedded

ADPCM Codec(Transmitter)

Fig. 5 Prediction and Inverse-quantized Sample Sign

Run-length Occurrence 9

7 ADPCM

ADPCM

Fig. 7 Con guration of the Proposed Text-Embedded

se(t)

ADPCM Codec(Receiver)

ADPCM

ADPCM i(t)

se(t)

55% ( 245**-*****)

48% ( 10828 5168)

ADPCM

dq (t)

ADPCM i(t)

[5] ADPCM

dq (t) 5 i(t)

se(t) dq (t)

6, 7

dq (t)

[4]

xxxx/xx Vol. Jxx D II No. xx

Table 1 Minimum Run Length vs. Embedded Text

Bits

ADPCM i(t) Speech Text Minimum Run Embedded Bits

Sample Type Length (Text Ratio

I 24,516 (9.56)

ADPCM II 1 10,828 (4.22)

I 10,850 (4.23)

Male II 2 5,580 (2.18)

dq (t)

I 5,574 (2.17)

II 3 4,208 (1.64)

I 873 (0.34)

ADPCM

II 10 2,178 (0.85)

I 26,375(10.39)

II 1 10,650 (4.16)

ADPCM i(t)

I 11,777 (4.60))

dq (t) Female II 2 6,122 (2.39))

I 6,529 (2.55))

II 3 4,781 (1.87)

I 867 (0.34)

II 10 2,728 (1.06)

Type I

Type II

2. 3

dq kmin Type I

[8]

ATR

se kmin Type

16kHz 16bit

II 3 Type II

8kHz

Type I

2. 3. 1

Type

ADPCM

2. 3. 2

se kmin se kmax

dq kmin, dq kmax

1 Type

Type I

2. 3. 3

dq kmin Type II

se kmin dq kmax se kmax 8 ADPCM

250 se(t) dq (t) Type I II

. dq kmin =

1 se kmin = 1, dq kmax = se kmax = 250

3 (Bit Error

1 Type I Type Pattern 1, 2, 3)

II 2.5 2

ADPCM

" #

" #$ # !"

$% & ' ) * *

(

# !"

>" #$

9 ADPCM

8 ADPCM

Fig. 9 Sign Bit Error Resilience of the ADPCM

Codec

Fig. 8 Bit Error Resilience of Embedded Text for the

ADPCM Codec

(Substituted Bits) !

(Inserted Bits)

(Deleted Bits) 9

Substituted Bits + Deleted Bits + Inserted Bits

BER [%] = 100

Total Evaluated Bits

Type II 9

Type I

10 ADPCM

9 Fig. 10 Con guration of the Proposed Dual Text-

Embedded ADPCM Codec(Transmitter)

ADPCM

Type I

ADPCM

Type

10 11

Type I Type

2. 4

Type II 3 Type

xxxx/xx Vol. Jxx D II No. xx

! "# $

' #

(&#

! ' # $

& # %#

! "# $

) #

%"#

12 ADPCM

Fig. 12 Con guration of the Sub Band ADPCM

!" #"

Codec (Transmitter)

11 ADPCM $

Fig. 11 Con guration of the Proposed Dual Text-

Embedded ADPCM Codec(Receiver) !

13 ADPCM

Fig. 13 Con guration of the Sub Band ADPCM

Codec (Receiver)

II se(t)

Type I

dq (t) 2

se(t) dq (t) Type I, II

Type

Type I Type I

Type II

3. 1 ADPCM

Type I Type II 12 13

Selector 7kHz G.722

Type [6] G.722 2

ADPCM

ADPCM 2bit/sample 6bit/sample

i(t) ADPCM

dq (t)

Type I dq (t)

ADPCM se(t)

se(t) Type

II ADPCM 3. 2 ADPCM

se(t) dq (t)

Type I

Type I ADPCM

se(t), dq (t) 2bit

6bit

ADPCM

3. 3 !$

"# %

!"# %

ADPCM 7kHz

$ %

MPEG

(EBU) SQAM(Sound Qual-

ity Assessment Material) [9]

1 1 Oregon

14 ADPCM

Graduate Institute

(Type I) Segmental SNR

SpEAR(Speech Enhancement

Fig. 14 Embedded Text (Type I) Ratio vs. Segmen-

Assessment) [10] 1 tal SNR for the SB-ADPCM Codec

16kHz 16bit

3. 3. 1

se kmax, se kmin,

dq kmax, dq kmin

ADPCM

Type I Type II

3. 3. 2 15 ADPCM

(Type I) (MOS)

14 Type I

Fig. 15 Embedded Text (Type I) Ratio vs. Mean

segmental SNR Opinion Score for the SB-ADPCM Codec

dq kmax 100, dq kmin 1 Type I

3. 3. 3

16 ADPCM

Type I II

segmental .

SNR SB-ADPCM dq kmin = se kmin = 1, dq kmax =

100, se kmax = 250 Type II

15 (Mean Type I

Opinion Score) piano Type II

[8] ATR 10% ADPCM

xxxx/xx Vol. Jxx D II No. xx

! "$ % '

#! & *

#! &

L. Boney, A. H. Tew k, K. N. Hamdy Digital Wa-

[1]

! "! +

! " $+ + '

#,, *

termarks for Audio Signals, Proc. IEEE Interna-

! "# ! % '

"$ & *

tional Conf. on Multimedia Computing and Systems,

! "# ! % ' )*

"$ &

pp.473-480, Hiroshima, Japan, Jun. 1996.

! "# +

[2] 40, 32, 24, 16 kbit/s Adaptive Di erential Pulse

! "# + + '

"$,, *

Code Modulation (ADPCM), ITU-T Recommenda-

tion G.726, 1990.

[3] 5, 4, 3 and 2 Bit/sample Embedded Adaptive Dif-

ferential Pulse Code Modulation (ADPCM), ITU-T

Recommendation G.727, 1990.

PCM

[4]

vol.

38, no. 10, pp.2053-2061, Oct. 1997.

[5]

2A4, p.19, Aug. 2000.

[6] 7kHz Audio-Coding within 64 kbit/s, ITU-T Rec-

16 ADPCM

ommendation, 1988.

[7]

Fig. 16 Bit Error Resilience of Embedded Text for

the SB-ADPCM Codec

02-4-A3, Mar. 2003.

[8]

1991

[9] European Broadcasting Union, SQAM(Sound Qual-

4. ity Assessment Material) - Recordings for Subjective

Tests, April 1988.

[10] CSLU, Oregon Graduate Institute of Science and

Technology, Speech Enhancement Assessment Re-

ADPCM source (SpEAR) Database, http://ee.ogi.edu/NSEL,

beta release v1.0.

G.726 727 7kHz

xx xx xx

G.722

2 57 59

2 4

Media Technologies Laboratory, Member of

Technical Sta 11

IEEE

ADPCM

12 Van-

tan J College 13

2 IEE 59 3

IEEE

Abstract We investigated on a low-complexity digital data embedding method applicable to ITU-T

standards G.726, G.727 ADPCM and G.722 Sub Band ADPCM. We proposed a method which embeds text

data in the least signi cant bit of the ADPCM output code when the inverse-quantized di erence signal

sign bit run-length is within its predetermined range. The proposed embedded text was proven to be robust

to bit errors, making it suitable to embed data which need to remain intact, e.g. copyright information,

while its e ect on the objective as well as subjective speech quality was found to be negligible.

Key words ADPCM, 7kHz Audio Coding, G.726, G.722, Digital Watermarks

Contact this candidate