Matching Items (2)
Filtering by
- Genre: Masters Thesis
Description
There is a tremendous need for wireless biological signals acquisition for the microelectrode-based neural interface to reduce the mechanical impacts introduced by wire-interconnects system. Long wire connections impede the ability to continuously record the neural signal for chronic application from the rodent's brain. Furthermore, connecting and/or disconnecting Omnetics interconnects often introduces mechanical stress which causes blood vessel to rupture and leads to trauma to the brain tissue. Following the initial implantation trauma, glial tissue formation around the microelectrode and may possibly lead to the microelectrode signal degradation. The aim of this project is to design, develop, and test a compact and power efficient integrated system (IS) that is able to (a) wirelessly transmit triggering signal from the computer to the signal generator which supplies voltage waveforms that move the MEMS microelectrodes, (b) wirelessly transmit neural data from the brain to the external computer, and (c) provide an electrical interface for a closed loop control to continuously move the microelectrode till a proper quality of neural signal is achieved. One of the main challenges of this project is the limited data transmission rate of the commercially available wireless system to transmit 400 kbps of digitized neural signals/electrode, which include spikes, local field potential (LFP), and noise. A commercially available Bluetooth module is only capable to transmit at a total of 115 kbps data transfer rate. The approach to this challenge is to digitize the analog neural signal with a lower accuracy ADC to lower the data rate, so that is reasonable to wirelessly transfer neural data of one channel. In addition, due to the limited space and weight bearing capability to the rodent's head, a compact and power efficient integrated system is needed to reduce the packaged volume and power consumption. 3D SoP technology has been used to stack the PCBs in a 3D form-factor, proper routing designs and techniques are implemented to reduce the electrical routing resistances and the parasitic RC delay. It is expected that this 3D design will reduce the power consumption significantly in comparison to the 2D one. The progress of this project is divided into three different phases, which can be outlined as follow: a) Design, develop, and test Bluetooth wireless system to transmit the triggering signal from the computer to the signal generator. The system is designed for three moveable microelectrodes. b) Design, develop, and test Bluetooth wireless system to wirelessly transmit an amplified (200 gain) neural signal from one single electrode to an external computer. c) Design, develop, and test a closed loop control system that continuously moves a microelectrode in searching of an acceptable quality of neural spikes. The outcome of this project can be used not only for the need of neural application but also for a wider and general applications that requires customized signal generations and wireless data transmission.
ContributorsZhou, Li (Author) / Muthuswamy, Jitendran (Thesis advisor) / Sutanto, Jemmy (Thesis advisor) / Yu, Hongyu (Committee member) / Arizona State University (Publisher)
Created2012
Description
There have been multiple attempts of coupling neural networks with external memory components for sequence learning problems. Such architectures have demonstrated success in algorithmic, sequence transduction, question-answering and reinforcement learning tasks. Most notable of these attempts is the Neural Turing Machine (NTM), which is an implementation of the Turing Machine with a neural network controller that interacts with a continuous memory. Although the architecture is Turing complete and hence, universally computational, it has seen limited success with complex real-world tasks.
In this thesis, I introduce an extension of the Neural Turing Machine, the Neural Harvard Machine, that implements a fully differentiable Harvard Machine framework with a feed-forward neural network controller. Unlike the NTM, it has two different memories - a read-only program memory and a read-write data memory. A sufficiently complex task is divided into smaller, simpler sub-tasks and the program memory stores parameters of pre-trained networks trained on these sub-tasks. The controller reads inputs from an input-tape, uses the data memory to store valuable signals and writes correct symbols to an output tape. The output symbols are a function of the outputs of each sub-network and the state of the data memory. Hence, the controller learns to load the weights of the appropriate program network to generate output symbols.
A wide range of experiments demonstrate that the Harvard Machine framework learns faster and performs better than the NTM and RNNs like LSTM, as the complexity of tasks increases.
In this thesis, I introduce an extension of the Neural Turing Machine, the Neural Harvard Machine, that implements a fully differentiable Harvard Machine framework with a feed-forward neural network controller. Unlike the NTM, it has two different memories - a read-only program memory and a read-write data memory. A sufficiently complex task is divided into smaller, simpler sub-tasks and the program memory stores parameters of pre-trained networks trained on these sub-tasks. The controller reads inputs from an input-tape, uses the data memory to store valuable signals and writes correct symbols to an output tape. The output symbols are a function of the outputs of each sub-network and the state of the data memory. Hence, the controller learns to load the weights of the appropriate program network to generate output symbols.
A wide range of experiments demonstrate that the Harvard Machine framework learns faster and performs better than the NTM and RNNs like LSTM, as the complexity of tasks increases.
ContributorsBhatt, Manthan Bharat (Author) / Ben Amor, Hani (Thesis advisor) / Zhang, Yu (Committee member) / Yang, Yezhou (Committee member) / Arizona State University (Publisher)
Created2020