Fused floating point arithmetic for discrete wavelet transform

Floating point arithmetic provides high resolution and wide range for representing numbers. However it requires huge resources and high power consumption which makes it unfavorable for implementation in power constrained wearable devices. In this paper we present fused floating point arithmetic for realizing Discrete Wavelet Transform (DWT). Comparative analysis is performed to determine the advantages of fused floating point. Implementing DWT using fused floating point dot product required 22% less area than using conventional floating point arithmetic. Moreover the required power consumption is 67% less. DWT is chosen in the analysis since it is widely used in wearable ECG processors due to its inherent behavior to suppress noise.