Quantitative image quality analysis of a nonlinear spatio-temporalfilter

Digital temporal and spatial filtering of fluoroscopic image sequences can be used to improve the quality of images acquired at low X-ray exposure. In this study, we characterized a nonlinear edge preserving, spatio-temporal noise reduction filter, the bidirectional multistage (BMS) median filter of Arce (1991). To assess image quality, signal detection and discrimination experiments were performed on stationary targets using a four-alternative forced-choice paradigm. A measure of detectability, d', was obtained for filtered and unfiltered noisy image sequences at different signal amplitudes. Filtering gave statistically significant, average d' improvements of 20% (detection) and 31% (discrimination). A nonprewhitening detection model modified to include the human spatio-temporal visual system contrast-sensitivity underestimated enhancement, predicting an improvement of 6%. Pixel noise standard deviation, a commonly applied image quality measure, greatly overestimated effectiveness giving 67% improvement in d'. We conclude that human testing is required to evaluate the filter effectiveness and that human perception models must be improved to account for the spatio-temporal filtering of image sequences.     

A silicon basis for synaptic plasticity

This paper describes analogue VLSI synaptic circuitry incorporating two forms of plasticity: a nonassociative mechanism analogous to paired-pulse facilitation (PPF), and an associative form of adaptation analogous to the short-term potentiation (STP) variety of plasticity. Simulation results demonstrate expected temporal characteristics of the plasticity mechanisms. The time-course of decay of associative potentiation is established by utilising diode leakage current.     

A Micropower Adaptive Linear Transform Vector Quantiser

Micropower signal classification and compression are becoming important requirements for implantable cardioverter defibrillators although they are currently limited by power and computational constraints. This paper describes an integrated circuit that facilitates Intracardiac Electrogram (ICEG) classification and compression of 30 dimensional analogue vectors while consuming a maximum of 2.5 W power for a heart rate of 60 beats per minute (1 vector per second) from a 3.3 V supply. This represents a significant advance on previous work which achieved ultra low power supervised morphology classification [6] since the templated matching scheme used in this chip enables unsupervised blind classification of abnormal rhythms and the computational support for low bit rate data compression. The adaptive template matching scheme used is tolerant to amplitude variations, and inter- and intra-sample time shifts. Micropower performance is achieved using CMOS analogue circuits biased in weak inversion in order to minimise energy per computation. Results from the fabricated chip demonstrate the impact of the amplitude and shift tolerance on ICEG data and performance for blind classification of an abnormal rhythm in five heart patients. For four out of the five patients, no false negative classifications and a worst case of 11% false positive classifications were made.     

The 3G-324M protocol for conversational video telephony

As mobile operators worldwide migrate to third-generation (3G) networks, conversational video-telephony services are becoming a key differentiator between new 3G offerings and existing 2G/2.5G services. Although it's possible to have limited video-based services - such as a multimedia messaging service - that deliver pictures and video clips over 2.5G services, these are delay-insensitive applications that could run over a packet-based wireless network like general packet radio service (GPRS) or code division multiple access (CDMA)'s 1XRTT. For delay-sensitive applications such as conversational video telephony, present 3G packet bearers are inadequate, and the Third Generation Partnership Project (3GPP; http://www.3gpp.org) mandates using the 3G bandwidth-guaranteed circuit-switched bearer and the 3G-324M system. The 3G-324M system is a derivative of the International Telecommunication Union (ITU) H.324 protocol standard for low-bitrate multimedia communication, which ITU-T developed for the public switched telephone network (PSTN). This article describes the 3G-324M system, which has been adopted by both 3GPP and 3GPP2 (htpp://www.3gpp2.org), as well as its H.324 roots.     

Compact low-power time-based conversion with noise immunity similar to digital conversion

An analog signal representation based on the inter-pulse-interval (IPI) time is presented. Voltage-to-IPI and IPI-to-voltage conversion circuits based on the representation are described. The circuits have been fabricated using a 0.35 μm mixed-signal CMOS process. Simulation and test results agree with the theory. Voltage-to-IPI conversion needs significantly less area and power than ADC and is significantly more immune to noise and other problems than using analog voltage/current signals.     

Parallel Fiber Coding in the Cerebellum for Life-Long Learning

Continuous and real-time learning is a difficult problem in robotics. To learn efficiently, it is important to recognize the current situation and learn appropriately for that context. To be effective, this requires the integration of a large number of sensorimotor and cognitive signals. So far, few principles on how to perform this integration have been proposed. Another limitation is the difficulty to include the complete contextual information to avoid destructive interference while learning different tasks.We suggest that a vertebrate brain structure important for sensorimotor coordination, the cerebellum, may provide answers to these difficult problems. We investigate how learning in the input layer of the cerebellum may successfully encode contextual knowledge in a representation useful for coordination and life-long learning. We propose that a sparsely-distributed and statistically-independent representation provides a valid criterion for the self-organizing classification and integration of context signals. A biologically motivated unsupervised learning algorithm that approximate such a representation is derived from maximum likelihood. This representation is beneficial for learning in the cerebellum by simplifying the credit assignment problem between what must be learned and the relevant signals in the current context for learning it. Due to its statistical independence, this representation is also beneficial for life-long learning by reducing the destructive interference across tasks, while retaining the ability to generalize. The benefits of the learning algorithm are investigated in a spiking model that learns to generate predictive smooth pursuit eye movements to follow target trajectories.     

Algorithmic and implementation issues in analog low power learning neural network chips

Analog sub-threshold is an attractive microelectronic implementation approach for many applications where power is to be minimized. The process of mapping a neural network to a sub-threshold architecture requires the proper selection of a training algorithm. For analog architectures, conventional training algorithms like backpropagation have many drawbacks although they are computationally efficient on digital computers. In this article we present algorithms that are suitable for analog implementation and we present architectures and implementations of sub-threshold neural networks.