ABSTRACT
The electronic skin system usually hosts many sensors coupled to an embedded electronic computing unit (ECU). This chapter presents the implementation of approximate arithmetic circuits in two blocks of the ECU: signal processing and data decoding. Usually, approximate results are acceptable for many applications, e.g., tactile data processing, image processing, and data mining; thus, it is possible to take advantage of energy reduction with minimal variation in performance. The chapter introduces the state-of-the-art low power low-pass finite impulse response (FIR) filters for DSP applications and shows the implementation results of a case study dealing with several approximate FIR filters. Quantization is used as a hardware approximation technique: the floating-point operations are substituted by fixed point ones. The chapter describes the tensorial kernel approach. It discusses the improvements obtained in terms of energy consumption and time latency after applying approximate computing techniques to the most expensive arithmetic circuits.
