J. L. Flanagan - jlf@caip.rutgers.edu
Rutgers University
96 Frelinghuysen Rd.
Piscataway, NJ 08854-8088
Popular version of paper 2pAAd1
Presented Tuesday afternoon, May 25, 2004
147th ASA Meeting, New York, NY
Communication acoustics has been a central theme in Bell Labs research since its inception. Telecommunication serves human information exchange. Humans favor spoken language as a principal mode. The atmospheric medium typically provides the link between articulation and hearing. Creation, control and detection of sound, and the human facility for generation and perception, are basic ingredients of telecommunication. Electronics technology of the 1920's ushered in great advances in communication at a distance, a strong economical impetus being to overcome bandwidth limitations of wireline and cable. Early research established criteria for speech transmission with high quality and intelligibility. These insights supported exploration of means for efficient transmission - obtaining the greatest amount of speech information over a given bandwidth. Transoceanic communication was initiated by undersea cables for telegraphy. But these long cables exhibited very limited bandwidth (order of few hundred Hz). The challenge of sending voice across the oceans spawned perhaps the best known speech compression technique of history - the Vocoder, which parameterized the signal for transmission in about 300 Hz bandwidth, one-tenth that required for the typical waveform channel. Quality and intelligibility were grave issues (and they still are). At the same time parametric representation offered possibilities for encryption and privacy inside a traditional voice bandwidth. Confidential conversations between Roosevelt and Churchill during World War II were carried over high-frequency radio by an encrypted vocoder system known as Sigsaly.
Major engineering advances in the late 40's and early 50's moved telecommunications
into a new regime - digital technology. These key advances were at least three:
(i) new understanding of time-discrete (sampled) representation of signals,
(ii) digital computation (especially binary based); and (iii) evolving capabilities
in microelectronics, that ultimately provided circuits of enormous complexity
with low cost and power. Digital transmission (as exemplified in Pulse Code
Modulation - PCM, and its many derivatives) became a telecommunication mainstay,
along with switches to control and route information in digital form. Concomitantly,
storage means for digital information advanced, providing another impetus for
speech compression. More and more, humans saw the need to exchange speech information
with machines, as well as with other humans. Human-machine speech communication
came to full stride in the early 1990's, and now has expanded to multimodal
domains that begin to support of enhanced naturalness, using contemporaneous
sight, sound and touch signaling. Traditional voice-band circuit switching is
being supplanted by high-speed packet transmission and Internet protocol. And,
heady mixes of multimodal information are beginning to support user-friendly
virtual environments that offer three-dimensional realism.