Performed By

A project seminar on performance analysis & implementation of AMR- WIDEBAND SPEECH CODEC PERFORMED BY (1) Kapadia Kunal R. (6229) (3) Desai Jaimin B. (6212)

(2) Desai Chintan B. (6211) (4) Chopra Dhikant R. (6208) GUIDED BY

(Prof. Ninad S. Bhatt) (Mr. Vijayendra A. Desai) ELECTRONICS & COMMUNICATION ENGINEERING DEPARTMENT OF C.K.PITHAWALA COLLEGE OF ENGINEERING & TECHNOLOGY,

 Problem statement:To design & implement AMR-WB.  We are going to implement AMR WIDEBAND USING MATLAB.  We have started our work with AMR & VAD.  IN our project we will used following ALGORITHM : I. LPC (LINEAR PREDICTIVE CODING) II. CELP (CODE EXITED LINEAR PREDICTIVE CODING) III. ACELP (ALGEBRIC CODE EXITED LINEAR PREDICTIVE CODING )

 SPEECH CODING speech coding is a procedure to represent a digitized speech signal using as few bits as possible, maintaining at the same time a reasonable level of speech quality.

SPEECH CODERS

according to bit rate High bit rate >15kbps Medium bit rate 5 to 10kbps Low bit rate 2 to 5kbps Very low bit rate < 2kbps

according to coding technique Waveform coders Parametric coders Hybrid coders

Requirements of speech coders Low bit rate High speech quality Robustness across different speakers Robustness in the presence of channel error Good performance on non speech signals Low memory size & low computational complexity Low coding delay

Basic block diagram of speech coding :

SPEECH

FILTER

SAMPLER

A/D CONVERT ER

SOURCE ENCODE R

CHANN EL ENCOD ER

CHANNEL

CHANNEL DECODER

SOURCE DECODE R

D/A CONVETER

FILTER

OUTPUT SPEECH

Limitation of other basic speech coders  SORCE CODER & CHANNEL CODER WORK SEPARETELY.

Solution

IN AMR, THIS LIMITATION OVERCOME THE PROBLEM OF SOURCE CODER & CHAN CODER WORKING SIMULTANEOUSLY.

 AMR(ADAPTIVE MULTI RATE ):  AMR is a technique to maintain good voice quality in various conditions of channel.  It is a technique which increases voice quality and boosted channel error robustness and capacity.

 Why AMR is used?  IN CELLULAR COMMUNICATION, quality of voice is most important. It depends on channel condition. Bad channel condition produces many error. TO maintain voice quality in different channel condition AMR is required.

BASIC DIAGRAM OF AMR CODEC SYSTEM : 



OPERATION OF AMR :

 AMR FULL RATE:[BIT RATE 22.8kbps] Total 8 modes of operation

 AMR HALF RATE:[BIT RATE 11.4kbps] Total 6 modes of operation

 From 12.2 kbps to 4.75kbps  From 7.95 kbps to 4.75kbps  In worst channel conditio:4.75kbps  In worst channel condition:4.75kbp  In best channel condition: 12.2kbps  In best channel condition: 7.95kbps

LINK ADAPTION & IN-BAND SINGALLING:  

IT CONSISTS OF CHANNEL QUALITY MEASUREMENT & CODEC MODE ALGORITHM. IT ADAPTS BIT PARTITIONING BETWEEN CHANNEL & SPEECH CODING. DEPENDING ON C/I, LINK ADAPTION SELECT THE OPTIMAL CODEC & CHANNEL MODE. C/I AMR (dB) MODE(kb/s ) 8

>=

7.95

>=5 <5

 AMR CONTAINS VAD & DTX.

12.2

5.90

VAD (VOICE ACTIVITY DETECTION): 

IT IS A TECHNIQUE USED IN SPEECH PROCESSING WHEREIN THE PRESENCE OR ABSENCE OF HUMAN SPEECH IS DETECTED IN REGION OF AUDIO.  Features of VAD:  LINE SPECTRAL FREQUENCIES  FULL-BAND ENERGY  LOW-BAND ENERGY(<1 kHz),  ZERO CROSSING RATE

 VAD application: AUDIO CONFERENCING  ECHO CANCELLATION SPEECH RECOGNITION SPEECH ENCODING HANDS-FREE TELEPHONY

 BLOCK DIAGRAM VAD FILTER ALGORITHM: S(i)

T_OP[n]

T0,t1

OL-LTP Correlati on vector

BANK & COMPUTATI ON OF SUB BAND LEVELS PITCH DETECTION

TONE DETECTION

COMPLEX SIGNAL ANALYSIS

level[n ]

pitch VAD DECESION

ton e

VAD FLAG

Complex warning Complex timer



DESIGN OF VAD ALGORITHM:

A. NOISE REDUCTION, e.g. via spectral subtraction. B. FEATURES OR QUANTITIES ARE CALCULATED FROM A SECTION OF SIGNAL INPUT. C. DETECT WHETHER SPEECH IS TRANSMITTED OR NOT.  OVERVIEW OF VAD BLOCK DIAGRAM:  FILTER BANK : Input signal is divided into frequency band using a 9 band filter bank. PITCH DETECTION: IT detects vowel sounds & other periodic signal.  TONE DETECTION: IT detects information tone.  COMPLEX SIGNAL ANALYSIS: It detects correlated signals in the high pass filtered weighted speech domain.  VAD DECESION: After comparing with threshold value it gives VAD output.



AMR WIDEBAND ADAPTIVE MULTI RATE (AMR-WB): It is a patented A speech coding standard developed after the AMR. Using similar technology ACELP.

 CHARACTERISTICS: Frequency range: 50Hz - 7000Hz  Sampling rate: 16KHz  Bit rate: 6.6-23.85Kb/s  APPLICATIONS:     

HIGH-FIDELITY TELEPHONY OVER BROABAND PACKET & ISDN THIRD-GENERATION MOBILE COMMUNICATION SYSTEM AUDIO & VIDEO TELECONFERENCING INTERNET APPLICATIONS DIGITAL RADIO BROADCASTING

REFERENCES: BOOK SUPPORT • Speech Coding Algorithms Foundation and Evolution of Standardizatied Coders by Wai C. Chu,A John Wiley & Sons,2003 • Wireless Communication Principle and Practice by Theodore S. Rappaport, 2005

WEB SUPPORT: • En.wikipedia.org • www.vocal.com • www.3gpp.org

THANK YOU!!!!