Low -power silicon microphone and sound recognition system.

A low-power sound recognition system was fabricated to recognize short-period sounds. The system consists of two parts: a high-sensitivity polysilicon diaphragm condenser microphone and a low-power mixed signal sound recognition chip. The microphone uses a low-stress polysilicon diaphragm suspended above a p+ perforated back plate, and it is fabricated by a combination of surface and bulk micromaching techniques. Measured results show sensitivities of -42 to -33 dB (ref. to 1 V/Pa) for 2 mm-wide diaphragms with a bias of 13 V and -44 to -37 dB for 2.6 mm-wide diaphragms with a bias of 10 V. The sound recognition chip uses the analog filter banks with energy method, and is implemented using MOS analog and digital integrated circuit design techniques. The filter banks are separated according to the auditory critical bands from 0.2 kHz to 12 kHz, and designed by the continuous time transconductor-capacitor technique. The core chip area is about 1.1 x 1.2 mm2 and the die area is 1.9 x 1.9 mm2. The measured power consumption is less than 60 mW with a clock rate of 21 kHz. With a combination of the silicon microphone and the sound recognition chip, a low-power sound recognition system was implemented. The system was tested to recognize the spoken word zero from the numerical numbers, zero to nine, with good selectivity (11.88 dB rejection). This project involves the analysis, design, fabrication, and testing of the silicon microphone and the sound recognition chip.Ph.D.Applied SciencesElectrical engineeringMechanicsSystems scienceUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/132378/2/9963808.pd