Blind Source-Separation in Mixed-Signal Vlsi

Summary

---

This paper describes an implementation of the Kurtosis and InfoMax algorithms for an independent components analysis in mixed-signal CMOS. Our design uses on-chip calibration techniques and local adaptation to compensate for the effect of device mismatch in arithmetic blocks and analog memory cells. We use our design to perform two-input blind source-separation on mixtures of audio signals and mixtures of EEG signals. Our experiments show that the hardware implementation of InfoMax consistently separates the signals within a normalized reconstruction error of less than 10%, while the reconstruction error of Kurtosis varies between 25% and 60%, due to its higher sensitivity to device mismatch and input statistics. Each circuit has a settling time of 8 µs, occupies a die area of 0.016-0.022 mm^sup 2^ and dissipates 15-20 µW of power.

See the full content of this document

Extract

---

Blind Source-Separation in Mixed-Signal Vlsi

(ProQuest: ... denotes formulae omitted.)

1. Introduction

Independent Components Analysis (ICA) [1, 2] is a signal processing technique used to recover the original sources from unknown linear combinations of independent signals captured by spatially-distributed sensors. Algorithms for ICA adapt without an external reference, and are used to perform blind source-separation, feature extraction and blind deconvolution in a wide range of applications such as speech recognition, face classification, biomedical signal and image analysis, and data communications [3, 4, 5].

Despite the advantages described above, most algorithms for ICA pose computational requirements that make them unsuitable for portable, low-power applications. Low-power embedded processors are unable to provide the throughput required for computation and adaptation, while high-performance digital signal processors (DSPs) are too large and power-hungry [6]. Even custom- VLSI digital implementations can be unsuited in ultra low-power applications, mainly due to the size and power requirements of digital multipliers [7, 8]. For problems that require only moderate arithmetic resolution, analog and mixed-signal circuits offer an attractive tradeoff between performance, die area and power dissipation. Unfortunately, large-scale analog implementations of signal processing algorithms are extremely difficult to design due to inherent limitations caused by device mismatch, charge leakage, charge injection, s...

See the full content of this document

---