Chapter Five
5. Speech Coding in GSM
Speech coding is a process that compresses and encodes speech signals to efficiently transmit voice
data over limited bandwidth while maintaining intelligibility and quality. Standards like GSM use
various speech coding techniques to reduce data rates while preserving speech clarity. The speech
codecs transform the analog voice signal into a digital format suitable for wireless transmission.
5.1. Speech Coding Methods
Speech coding methods are primarily designed to achieve high compression while maintaining
acceptable voice quality. The main speech coding techniques include:
5.1.1. Wavefor m Coding
Waveform coding aims to reproduce the original speech waveform as accurately as possible using
pulse code modulation (PCM). Examples include:
Pulse Code Modulation (PCM): Used in traditional telephone networks (64 kbps).
Adaptive Differ ential PCM (ADPCM): A more efficient method using past samples to
predict future values.
5.1.2. Vocoder -Based Coding
Vocoders analyze speech characteristics and send only essential parameters. Examples include:
Linear Pr edictive Coding (LPC): Models speech based on past samples.
Code-Excited Linear Pr ediction (CELP): Uses a codebook of predefined sound patterns.
5.1.3. Hybr id Coding
Hybrid coding combines waveform and vocoder methods to improve efficiency and quality. GSM
codecs use hybrid coding approaches like:
Regular Pulse Excitation-Long-Ter m Pr ediction (RPE-LTP): Used in GSM Full Rate
codec.
Algebr aic Code-Excited Linear Pr ediction (ACELP): Used in Enhanced Full Rate (EFR)
codec.
5.2. Speech Codec Attr ibutes
Speech codecs used in GSM are evaluated based on several attributes:
Bit Rate
The rate at which speech data is transmitted. GSM codecs operate at different bit rates:
Full Rate (FR): 13 kbps
Half Rate (HR): 5.6 kbps
Enhanced Full Rate (EFR): 12.2 kbps
Adaptive Multi-Rate (AMR): Variable rates from 4.75 to 12.2 kbps
Speech Quality
Measured in terms of Mean Opinion Score (MOS), where listeners rate quality on a scale from 1
(bad) to 5 (excellent). AMR provides the best quality among GSM codecs.
Complexity
Refers to the computational requirements of the codec. More complex algorithms provide better
compression but require more processing power.
Delay
The time taken to encode and decode speech. Lower delay ensures better real-time communication.
Er r or Robustness
How well the codec handles errors due to network issues. AMR codecs adapt their bit rate to
improve robustness.
5.3. LDPAS (Low-Delay Code-Excited Linear Pr ediction
Analysis-by-Synthesis)
LDPAS is a speech coding technique that improves coding efficiency while reducing encoding and
decoding delays. It is a variant of Code-Excited Linear Prediction (CELP) that focuses on:
Low latency: Ensuring real-time communication with minimal delay.
Efficient compr ession: Maintaining voice quality at lower bit rates.
Er r or r esilience: Making speech transmission robust against packet losses in GSM
networks.
LDPAS is widely used in mobile and VoIP applications, as it balances quality, bit rate, and
processing complexity.
5.4. ITU-T Standar ds
The International Telecommunication Union - Telecommunication Standardization Sector (ITU-T)
defines global standards for speech coding used in GSM and other communication systems. Some
key ITU-T standards include:
G.711: PCM codec for 64 kbps (used in fixed-line telephony).
G.726: ADPCM codec for bit rates from 16 to 40 kbps.
G.729: Low-bit-rate codec (8 kbps) based on CS-ACELP, used in VoIP.
G.723.1: Low-bit-rate codec (5.3 and 6.3 kbps) used in internet telephony.
G.722: Wideband codec for high-quality speech (used in HD voice calls).
In the context of GSM, the Adaptive Multi-Rate (AMR) codec, which is standardized under ITU-
T G.722.2, is commonly used for modern mobile networks, providing better quality and efficiency.
5.5. Vocoder s
A vocoder (short for voice encoder ) is an audio processing technology used to analyze, synthesize,
and compr ess speech signals. It works by separating the speech signal into its char acter istics (e.g.,
pitch, tone) and content (formants or shape of the sound) and then reconstructing or modifying the
speech.
What Does a Vocoder Do?
A vocoder takes a human voice input, analyzes it, and encodes it into a digital signal. It can then
transmit this encoded signal and synthesize it back into speech or use it for special audio effects.
Key Functions:
Speech Compr ession – Reduces bandwidth for efficient transmission.
Speech Synthesis – Recreates human speech from encoded data.
Voice Tr ansfor mation – Alters voice pitch, gender, or texture (common in music and
entertainment).
Encr yption – Secures speech transmission by encoding voice signals.
How Does a Vocoder Wor k?
A vocoder operates in two main stages: analysis and synthesis.
(a) Analysis Stage:
1. Input Signal – Captures speech (modulator) and a carrier signal (e.g., a musical tone).
2. Bandpass Filter ing – Divides speech into multiple fr equency bands.
3. Envelope Detection – Measures the amplitude of each band (speech characteristics).
4. Featur e Extr action – Identifies pitch, for mants, and spectr al content.
(b) Synthesis Stage:
1. Car r ier Signal – A constant sound (e.g., a synthesizer tone) is modulated by speech
patterns.
2. Amplitude Modulation – The carrier is reshaped using the speech characteristics.
3. Output Signal – Produces a r obotic, synthesized, or compr essed version of the speech.
Types of Vocoder s
(a) Analog Vocoder
Uses analog filter s to process and synthesize speech.
Applications: Early r obotic voices in music (e.g., Kraftwerk).
Example: Moog and Roland analog vocoders.
(b) Digital Vocoder
Uses digital signal pr ocessing (DSP) for higher precision and better compression.
Applications: Mobile communications, speech synthesis.
Example: Linear Pr edictive Coding (LPC).
Vocoder s in Telecommunications
Vocoders are essential in cellular communication to compress voice for efficient transmission.
Common Speech Coding Standar ds Using Vocoder s:
Vocoder Type Technology Application
LPC (Linear Pr edictive Coding) Predicts speech samples VoIP, Speech Synthesis
CELP (Code-Excited Linear Combines LPC with
GSM, 2G, 3G networks
Pr ediction) codebooks
Dynamic bit-rate speech GSM, LTE (4G), Voice-over-LTE
AMR (Adaptive Multi-Rate)
coding (VoLTE)
High-fidelity, low-latency
EVS (Enhanced Voice Ser vices) 4G LTE, 5G Voice
speech
Advantages and Disadvantages of Vocoder s
Aspect Advantages Disadvantages
Efficiency Compresses speech for low-bandwidth Loss of speech natur alness
Ver satility Used in telecom, music, and security Requires complex pr ocessing
Secur ity Can encr ypt speech for secure channels Susceptible to noise and distortion
Cr eativity Enables voice tr ansfor mation effects Reduces voice clar ity in some cases
Applications of Vocoder s
◦ Telecommunications – Compressing voice signals (GSM, VoIP).
◦ Music Pr oduction – Creating robotic or altered voice effects.
◦ Speech Synthesis – Text-to-speech (TTS) systems.
◦ Militar y Communications – Secure voice transmission.
◦ Assistive Technologies – Synthetic speech for people with speech disabilities.