A Floating-gate Vector-quantizer

We present a floating-gate based system for computing vector quantization (VQ), which is typically used for data compression and classification of signals to symbols. We present an architecture and resulting circuits which will enable direct programming / storage of weight vectors, as well as methods for adaptive VQ. We use an analog bump circuit to perform a continuous distance computation along a particular input coordinate. Unlike a traditional bump circuit, we use differential floating-gate inputs to provide the ability to store the learned value. The current outputs of each bump circuit are summed along a single wire, where the largest result(s) are selected using a winner-take-all circuit. We present experimental results measured from ICs fabricated on a 0.5um CMOS process available through MOSIS. Figure 1 shows the basic concepts in vector quantization (VQ), as well as its application in speech recognition. VQ is typically used in data compression and in classifying signals to symbols [1]. For example in speech processing, VQ is used to reduce the set of detectable spectrum vectors to a manageable set for later classification. An Analog-to-Digital Conversion (ADC) classifies data in a single dimension; a vector quantizer classifies data in an arbitrary number of dimensions. VQ is typically used in data compression, and like an ADC, VQ results in lossy compression. The goal of VQ is to provide the simplest possible accurate description of a signal so as to minimize the subsequent complexity of signal processing algorithms such as classification, We present a floating-gate based system for computing vector quantization (VQ), and present experimental results measured from ICs fabricated on a 0.5um CMOS process available through MOSIS. An analog implementation of VQ provides a combination of computing higher input signal frequencies or requiring lower power levels than the equivalent digital VQ implementation. This computational efficiency is important for low-power applications, such as portable battery powered devices. We see this technology as a fundamental component in a low-power speech recognition system [2], as well as in other sensory processing applications. 1. MATHEMATICAL BASIS OF VQ A VQ system will compute how far away a particular input vector is from the desired target vectors, and pick the code vector that is closest to the input vector. For example for an ADC, we assign a particular digital code to a given input based upon which analog This work was partially supported by grants National Science Foundation ( CISE-1068549, ECS (CAREER): 0093915, ECS-9988905) and by corportate donations to the Georgia Tech Analog Consortium by Texas Instruments and Motorola, Inc. VQ Microphone (a)