edVirtus Reference

Who is Andrew Viterbi?

Andrew Viterbi (1935-): The Engineer Who Made Most-Likely Sequence Decoding Practical

Andrew James Viterbi is an Italian-born American communications engineer, educator, and entrepreneur whose 1967 algorithm finds a most likely path through a system with memory. Created for decoding Convolutional Code, the Viterbi algorithm became a standard method in satellite, mobile, modem, recording, speech, and biological sequence applications.

Viterbi also co-founded Linkabit and Qualcomm and contributed to commercial Code Division Multiple Access. His career demonstrates a recurring feature of communications: a mathematical method acquires its full importance when computing, integrated circuits, standards, and complete systems make it deployable.

From Bergamo to American Engineering

Viterbi was born Andrea Giacomo Viterbi in Bergamo on 9 March 1935. Anti-Jewish laws under Italian Fascism forced his family to leave in 1939. They settled in the United States, where his first name became Andrew.

He studied electrical engineering at MIT in the intellectual environment of Claude Shannon, Norbert Wiener, and Roberto Fano, then completed a doctorate at the University of Southern California in 1962. He joined UCLA and taught digital communication and information theory.

Codes with Memory

A convolutional encoder combines current input with a limited history held in shift-register state. Different input sequences trace different paths through a trellis, a diagram showing the permitted state transitions and output symbols over time.

Noise makes the received observations inconsistent with every ideal path. Decoding therefore asks which complete path best matches the evidence, not which isolated bit looks largest. An exhaustive search grows exponentially with message length and is unusable for sustained communication.

The Survivor Principle

Viterbi Decoding assigns a metric to each possible transition and accumulates metrics along trellis paths. When two paths enter the same state at the same time, their future possibilities are identical. The path with the worse accumulated metric can be discarded permanently.

Keeping one survivor per state prevents the number of candidates from growing with message length. After sufficient delay, the decoder traces back through survivor decisions to estimate the transmitted sequence. Complexity depends mainly on the number of encoder states, so large constraint length remains costly even though long messages do not cause exponential growth.

Maximum Likelihood and Soft Information

With an appropriate channel model and metric, the algorithm performs maximum-likelihood sequence estimation. It chooses the allowed sequence that makes the observed signal most probable, rather than promising that every noisy message will be recovered correctly.

A hard-decision decoder first converts each observation to a bit. A soft-decision decoder retains reliability information such as distance from a decision boundary, often providing useful Coding Gain. This illustrates how a smarter receiver can extract performance without changing the transmitted code.

From Reinterpretation to General Algorithm

Viterbi introduced the method as an asymptotic decoding technique, and later work clarified its dynamic-programming and shortest-path interpretation. Jim Omura supplied an important link to maximum-likelihood decoding, while other researchers developed efficient hardware architectures and variants.

The algorithm applies whenever observations arise from a finite-state process. Hidden Markov models use a closely related recursion, which explains applications in speech recognition, magnetic recording, and computational biology. The trellis need not represent a radio code; it represents constrained sequences and accumulated evidence.

Space Links and Digital Hardware

Satellite and deep-space links made strong Forward Error Correction valuable because received power is scarce and retransmission may be impractical. As integrated circuits improved, Viterbi decoders could process longer constraint lengths and higher data rates within realistic power and mass limits.

Convolutional coding and Viterbi decoding became standard components in modems, telemetry, digital broadcasting, and early cellular systems. Later Turbo Codes and Low-Density Parity-Check Codes offered stronger performance in many applications, but the Viterbi algorithm remains important for specific codes, channels, and constituent decoders.

Linkabit, Qualcomm, and CDMA

Viterbi co-founded Linkabit in 1968 with Irwin Jacobs and Leonard Kleinrock. The company translated coding and signal-processing theory into satellite and military communication products, creating both hardware and a community of engineers who later shaped wireless industry.

In 1985 Viterbi, Jacobs, and colleagues founded Qualcomm. Viterbi contributed theoretical expertise and leadership as the company developed Direct Sequence Spread Spectrum cellular CDMA, where coding, power control, multipath processing, and receiver algorithms jointly determine system capacity.

Education, Recognition, and Legacy

Viterbi later supported universities and research on a major scale; USC named its engineering school for Andrew and Erna Viterbi. He received the National Medal of Science for the maximum-likelihood algorithm and contributions to CDMA wireless technology.

His enduring insight is the safe elimination of alternatives. Once two partial paths reach the same state, only the better one can become the best complete path. That simple structural fact transformed a prohibitive search into practical sequence inference and taught generations of receivers how to make disciplined decisions under uncertainty.