AMR-NB: Difference between revisions

Revision as of 22:12, 22 June 2007

AMR-NB (Adaptive Multi-Rate Narrowband) is a vocoder employed in low-bitrate applications like mobile phones.

Specification (26.101) describes two possible frame types - interface formats 1 and 2 (often abbreviated IF1 and IF2). IF2 is byte-aligned.

bits	low level meaning	high level meaning
4	Frame type	AMR header
1	Frame quality indicator (0 bad/1 good)	AMR header
3	Mode indication	AMR auxiliary information
3	Mode request
8	CRC
	Class A bits	AMR core frame
	Class B bits
	Class C bits

bits	low level meaning	high level meaning
4	Frame type	AMR header
	Class A bits	AMR core frame
	Class B bits
	Class C bits
	Padding (called "Bit stuffing" in the specification)

Frame type	Frame content
0	AMR 4.75kbps
1	AMR 5.15kbps
2	AMR 5.90kbps
3	AMR 6.70kbps (PDC-EFR)
4	AMR 7.40kbps (TDMA-EFR)
5	AMR 7.95kbps
6	AMR 10.2kbps
7	AMR 12.2kbps (GSM-EFR)
8	AMR SID
9	GSM-EFR SID
10	TDMA-EFR SID
11	PDC-EFR SID
12-14	Reserved for future use
15	No data (no transmission/no reception)

Class	Importance explanation
A	Data that is most sensitive to error. Any error in these bits leads to a corrupted speech frame that should not be decoded without appropriate error concealment. This class of bits is protected by an 8-bit CRC.
B	Less sensitive data that are present in all frames.
C	Least sensitive data present only in higher bit rate frames.

Class A is protected by an 8-bit CRC with polynomial x^8+x^6+x^5+x^4+1 computed over the Class A bits.
There is no significant step-wise change in subjective importance at class boundaries.
The distribution of bits is ordered from most to least subjective importance at both the class level and within the classes.

@@ Line 18: / Line 18: @@
 |-
 | 1
-| Frame quality indicator
+| Frame quality indicator (0 bad/1 good)
 |-
 | 3
@@ Line 99: / Line 99: @@
 |}
-=== Notes ===
+=== Classes ===
-* Frame quality indicator bit: 0 bad/corrupted (can use error concealment), 1 good
+{| border=1
-* CRC (8 bits) - CRC with polynomial x^8+x^6+x^5+x^4+1 computed over Class A bits
+! Class !! Importance explanation
-* Class A bits - the most important data
+|-
-* Class B bits - less important data
+| A || Data that is most sensitive to error. Any error in these bits leads to a corrupted speech frame that should not be decoded without appropriate error concealment. This class of bits is protected by an 8-bit CRC.
-* Class C bits - additional not very important data that may be present only on higher bitrates (frame types 6 and 7)
+|-
-How to smear bits between all these classes is defined by a so-called 'importance function'
+| B || Less sensitive data that are present in all frames.
+|-
+| C || Least sensitive data present only in higher bit rate frames.
+|}
+* Class A is protected by an 8-bit CRC with polynomial x^8+x^6+x^5+x^4+1 computed over the Class A bits.
+* There is no significant step-wise change in subjective importance at class boundaries.
+* The distribution of bits is ordered from most to least subjective importance at both the class level and within the classes.
 [[Category:Vocoders]]
 [[Category:Audio Codecs]]
 [[Category: Formats missing in FFmpeg]]