edVirtus Reference

What Is a Vocoder?

What Is a Voice Coder?

Preview: Learn more about vocoders and how they represent speech using mathematical models rather than transmitting the speech waveform directly.

A vocoder (short for voice coder) is a speech-coding system that analyses the characteristics of human speech and transmits a mathematical description of how the speech was produced rather than the speech waveform itself. Instead of sending every speech sample, a vocoder models the human vocal tract and transmits only the parameters needed to recreate the speech at the receiver. This approach allows intelligible speech to be transmitted at very low bit rates, making vocoders widely used in military communications, satellite systems, digital radio, and secure voice applications.

The basic principle of a vocoder is based on the source-filter model of speech production. In this model, the lungs provide an airflow, the vocal cords generate an excitation signal, and the vocal tract acts as a time-varying acoustic filter that shapes the sound into recognizable speech. Rather than reproducing the speech waveform directly, the vocoder estimates the characteristics of this filter together with the excitation signal and transmits only those parameters.

At the receiver, a speech synthesizer reconstructs the speech by generating an artificial excitation signal and passing it through a digital model of the vocal tract. Because only a relatively small number of parameters need to be transmitted, vocoders require much lower bit rates than conventional waveform coders such as Pulse Code Modulation (PCM).

A useful analogy is describing how to play a musical instrument rather than transmitting a recording of the performance. Instead of sending every sound wave produced by the instrument, one sends the musical notes, timing, and playing style, allowing another musician to recreate the performance. Similarly, a vocoder transmits instructions describing how the speech was produced rather than the complete speech waveform.

The first practical vocoder was developed by the American engineer Homer Dudley at Bell Telephone Laboratories in the late 1930s. His system analysed speech into several frequency bands and transmitted only the slowly varying amplitudes of those bands together with information about the excitation source. Although the reconstructed speech sounded distinctly synthetic, it remained intelligible while requiring far less bandwidth than conventional speech transmission.

During the Second World War, vocoder technology formed the basis of the highly secret SIGSALY secure speech system, one of the earliest digital encryption systems. Since then, vocoder technology has evolved dramatically through techniques such as Linear Predictive Coding (LPC), Adaptive Predictive Coding (APC), Regular Pulse Excitation (RPE), Multipulse Excitation (MPE), and Code Excited Linear Prediction (CELP). Modern speech codecs used in mobile telephones and Voice over IP (VoIP) systems continue to build upon these same modelling principles.

It is important to distinguish a vocoder from a waveform coder. Waveform coders attempt to reproduce the original speech waveform as accurately as possible, whereas vocoders seek to reproduce the speech-production mechanism. As a result, vocoders can operate at much lower bit rates, although the reconstructed speech may sound more synthetic than speech produced by high-rate waveform coders.

Today, vocoder technology underpins many low-bit-rate speech communication systems. Military radios, satellite communication systems, secure telephony, digital land-mobile radio, and modern speech codecs all employ techniques derived from the original vocoder concept. Although the quality of modern systems has improved enormously, the underlying principle remains unchanged: represent speech by transmitting its essential characteristics rather than every individual sample.