Efficient narrow-band speech transmission over land mobile channels.

This thesis addresses the problems of speech quality, and channel bandwidth efficiency, for speech transmission in mobile communications.In the first part, a flexible, real-time, programmable, mobile radio channel simulator is proposed. The conventional basic configuration of the simulator was augmented with several blocks, allowing the simulation of multipath fading, Doppler shifts, line of sight propagation and shadowing, with a wide range of parameters. This was achieved through the use of variable rate and multirate digital signal processing, combined with an appropriate system design allowing an efficient task partition between software and hardware. The simulator was built and evaluated. Its output was compared to several field tests; a good agreement is demonstrated.In the second part, a system is proposed for reducing the background acoustic noise from the transmitted speech signal in mobile telephony, and, therefore, improve the signal to noise ratio at the transmitter. Several field tests were conducted in order to characterize the background acoustic noise in cars. Accordingly, an adaptive noise reduction system is porposed, using a transversal finite impulse response adaptive filter and the stochastic gradient adaptation algorithm; a 10noise reduction is achieved.In the third part, a novel concept, the switched tone (), is proposed for speech and data transmission over mobile channels. The speech is transmitted using a modified single sideband (SSB) technique with a channel separation of 5kHz, instead of the 20-30kHz currently used in analog frequency modulation (FM).Three main subsystems, required by theconcept, are proposed. A novel differential dynamic speech companding system improves the signal to noise ratio of the low volume syllables, and, therefore, improves the speech quality. An appropriate Automatic Gain Control (AGC) system compensates for the amplitude fluctuations attributed to multipath fading. The tracking range of the AGC is programmable and can take any value according to the signal to noise ratio encountered. Finally, the Automatic Frequency Control (AFC) problem is solved in a systemwise approach through theconcept. The problem is reduced to two simpler ones, the initial acquisition, performed by a mid-band tone, and the phase tracking, performed by a tone above the speech band, allowing a full transparency to all speech and data formats.