Voice Over IP Introduction to digital voice technology
Digitizing Analog Signals
Analog signal value are expressed in term of Amplitude, Frequency and Phase.
Analog input is sent to Codec, output is digital stream of 0’s and 1’s.
Digitizing Analog Signals
Sampling. Quantization. Encoding. Compression (Optional).
Sampling Nyquist Theorem In order to digitize an analog signal, the signal must be sampled at a rate equal to that of twice the highest frequency of the signal to be digitized.
Sampling Audible spectrum of frequencies 200 Hz
20,000 Hz Human voice frequencies
250 Hz
10,000 Hz “Typical” Telephone channel frequencies
300 Hz
3,000 Hz
Sampling
For Telephone Voice = 300 ------ 3000 Hz
Because of Variation = 270 ------ 3300 Hz
Nyquist
=0
------ 4000 Hz
Sampling Nyquist Theorem Fq = 2fm Fq = 2 (4000) Hz Fq = 8000 Hz
Sampling ts = 1 / 8000 ts = 125 ms So take a sample after every 125 Micro Sec.
Sampling
Sampling Sample 125 ms
Sample 125 ms
Sampling G.711 1 sample = 8 bits 8000 Sample = 64000 = 64 kbps G.729 1 sample = 1 bits 8000 Sample = 8000 = 8 kbps
Quantization
Express a mathematical value for each of the Sample taken. In the case of digitizing analog voice, the values assigned are in the form of 8-bits binary words.
Quantization
256 combinations 255 are used 127 above zero, 127 below and 1 for zero it self. Bit patern of all zero’s is never used. mu-Law a-Law
Quantization Amplitude Quantization scale Example: this sample maps to this quantifiable value
Quantified value (8- bit values assigned to each delineation)
Zero reference
Time
Note: Each line represents 1/8000 of a second (samples)
Quantization Amplitude Quantization scale Example: this sample maps to this quantifiable value
Quantified value (8- bit values assigned to each delineation)
Zero reference
Time
Note: Each line represents 1/8000 of a second (samples)
Quantization
If a samples amplitude fell exactly between two samples – “quantization error” Quantization error creates “quantization noise” One error may not be heard but continuous error can create hearable noise.
Encoding
Sampling
Quantization
Encoding
Represents values in 8 bits reference.
Encoding P
Se
Se
Se
P - Polarity Se - Segment St - Step
St
St
St
St
Encoding
First bit represents Polarity (+ve , -ve) Next 3 bits represents Segments, sometimes called Chord, the area of the scale in which this sample is found. Next 4 bits represents Steps on delineation scale.
Encoding 1
0 P
1 Se
1
0
1
0
1
St
1st bit shows that Polarity is above zero. “3” (011) shows that sample situated in 3rd segment. “5” (0101) shows sample is on the 5th delineation within the segment 3.
Review Sampling. Samples are taken at a rate of 8000 t/s. Quantization. Each sample is quantified in comparison to a scale that has delineation grouped in segments
Review Encoding. Each quantifies sample will produce an encoded 8-bit word that represents the sample’s amplitude. NOTE: 8000 samples per second times 8 bits per sample yields 64000 - bits per second to represent one second of sound.
Compression
Wave Form Compression: Fellows the approach used for PCM encoding.
Vocoder Compression: Synthesized voice with processing intelligence.
Hybrid Compression: A combination of wave and vocoder compression.
Wave Form Compression
Subset of the encoding schemes.
Known as “Wave Form Coding”.
Tracks and fellows the actual wave form as it develops in real time. Comparative Differential Values.
Wave Form Compression
ADPCM
40000, 32000, 24000, 16000 bps
ITU-T G.726 and G.727
Wave Form Compression
Pros – Reduced bandwidth consumption Simple and inexpensive to process
Cons – Poorer audio quality at low rate.
Vocoder Compression
Pros – Deduced bandwidth consumption
Cons – Expensive to process Requires specialized electronics Sound synthetic Speaker is not recognized
Hybrid Compression
Combines best of Wav and Vocoder compression techniques High quality voice at low bit rate Extensively used in the digital cellular telephone industry.
Hybrid Compression
Coded Excited Linear Predictive ( CELP ) Has a “Code Book” Like STAC
Low Delay CELP (LD - CELP ) Makes a Code Book directly from speaker’s ITU-T’s G.728 at 16000 bps
voice
Hybrid Compression
Conjugate Structure Algebraic CELP Modified LD – CELP High – Quality Speech at 8000 bps
Hybrid Compression
Pros – Excellent audio quality Very low bit rates Adapts to speaker
Cons – Requires specialized processing chips Requires memories Induces processing delay
Digital Speech Interpolation
Voice Activity Detection ( VAD )
Digital Speech Interpolation (DSI) is a DSP function
“Enter Silent Period”
“Comfort noise”