Neuro-Inspired Speech Recognition Based on Reservoir Computing

Ghani, Arfan; McGinnity, TM; Maguire, Lisa; McDaid, Liam; Belatreche, Ammar

doi:10.5772/10186

Cited by 14 publications

(14 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As powerful as the brain, the neuromorphic computing system potentially solves computing-intensive tasks that are only handled by the human brains before. These multifaceted tasks include speech recognition [4][5][6], character recognition [7,8], grammar modeling [9], noise modeling [10], as well as the generation and prediction of chaotic time series [11,12], etc. However, state-of-the-art neuromorphic chips with the traditional CMOS technology and the two-dimensional (2D) design methodology cannot meet the energetic and speed requirements at large-scale neuron and synapse realization [13][14][15][16][17].…”

Section: Introductionmentioning

confidence: 99%

The Roadmap to Realize Memristive Three-Dimensional Neuromorphic Computing System

An¹,

Bai²,

Yi³

2018

Advances in Memristor Neural Networks - Modeling and Applications

View full text Add to dashboard Cite

Neuromorphic computing, an emerging non-von Neumann computing mimicking the physical structure and signal processing technique of mammalian brains, potentially achieves the same level of computing and power efficiencies of mammalian brains. This chapter will discuss the state-of-the-art research trend on neuromorphic computing with memristors as electronic synapses. Furthermore, a novel three-dimensional (3D) neuromorphic computing architecture combining memristor and monolithic 3D integration technology would be introduced; such computing architecture has capabilities to reduce the system power consumption, provide high connectivity, resolve the routing congestion issues, and offer the massively parallel data processing. Moreover, the design methodology of applying the capacitance formed by the through-silicon vias (TSVs) to generate a membrane potential in 3D neuromorphic computing system would be discussed in this chapter.

show abstract

Section: Introductionmentioning

confidence: 99%

The Roadmap to Realize Memristive Three-Dimensional Neuromorphic Computing System

An¹,

Bai²,

Yi³

2018

Advances in Memristor Neural Networks - Modeling and Applications

View full text Add to dashboard Cite

show abstract

“…This leads to decaying transient memories represented by the dynamic response of the reservoir to input spike trains. For this reason, the LSM is specially competent for processing temporal patterns such as speech signals [19], [20]. For readout neurons, Verstraeten et al [19] used ridge regression to calculate synaptic weights between the reservoir and readout neurons for classification tasks.…”

Section: Introductionmentioning

confidence: 99%

“…For readout neurons, Verstraeten et al [19] used ridge regression to calculate synaptic weights between the reservoir and readout neurons for classification tasks. Ghani et al [20] trained output neurons using backpropagation-based multilayer perceptrons.…”

Section: Introductionmentioning

confidence: 99%

“…Furthermore, the LSM is computationally powerful such that its performance on isolated word recognition can be comparable to state-of-art signal processing methods. Compared with other neural-network-based methods, such as Long Short-Term Memory nets [24] and multilayer perceptron-based classifiers [20], the LSM is either superior in performance or more hardware friendly and biologically plausible. However, the offline learning in [19] required large amounts of storage for intermediate results during training.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Digital Liquid State Machine With Biologically Inspired Learning and Its Application to Speech Recognition

Zhang

Jin

et al. 2015

IEEE Trans. Neural Netw. Learning Syst.

150

111

View full text Add to dashboard Cite

This paper presents a bioinspired digital liquid-state machine (LSM) for low-power very-large-scale-integration (VLSI)-based machine learning applications. To the best of the authors' knowledge, this is the first work that employs a bioinspired spike-based learning algorithm for the LSM. With the proposed online learning, the LSM extracts information from input patterns on the fly without needing intermediate data storage as required in offline learning methods such as ridge regression. The proposed learning rule is local such that each synaptic weight update is based only upon the firing activities of the corresponding presynaptic and postsynaptic neurons without incurring global communications across the neural network. Compared with the backpropagation-based learning, the locality of computation in the proposed approach lends itself to efficient parallel VLSI implementation. We use subsets of the TI46 speech corpus to benchmark the bioinspired digital LSM. To reduce the complexity of the spiking neural network model without performance degradation for speech recognition, we study the impacts of synaptic models on the fading memory of the reservoir and hence the network performance. Moreover, we examine the tradeoffs between synaptic weight resolution, reservoir size, and recognition performance and present techniques to further reduce the overhead of hardware implementation. Our simulation results show that in terms of isolated word recognition evaluated using the TI46 speech corpus, the proposed digital LSM rivals the state-of-the-art hidden Markov-model-based recognizer Sphinx-4 and outperforms all other reported recognizers including the ones that are based upon the LSM or neural networks.

show abstract

“…There have been several studies in the past to investigate the paradigm of RC [3, 8, 9] but none of them provide any guidelines as how to implement and analyse a stable reservoir on hardware/software (HW/SW) platforms. To address this deficiency, authors in [10, 11] demonstrated the viability of implementing neural reservoirs on software platforms. The main focus of this research was to investigate and analyse the impact of input connectivity and to elaborate the parameters that affect the stability of neural reservoirs.…”

Section: Introductionmentioning

confidence: 99%

Step forward to map fully parallel energy efficient cortical columns on field programmable gate arrays

Ghani

See

Ali

2014

IET Science, Measurement & Technology

Self Cite

View full text Add to dashboard Cite

This study presents energy and area-efficient hardware architectures to map fully parallel cortical columns on reconfigurable platform -field programmable gate arrays (FPGAs). An area-efficient architecture is proposed at the system level and benchmarked with a speech recognition application. Owing to the spatio-temporal nature of spiking neurons it is more suitable to map such architectures on FPGAs where signals can be represented in binary form and communication can be performed through the use of spikes. The viability of implementing multiple recurrent neural reservoirs is demonstrated with a novel multiplier-less reconfigurable architectures and a design strategy is devised for its implementation.

show abstract

Neuro-Inspired Speech Recognition Based on Reservoir Computing

Cited by 14 publications

References 36 publications

The Roadmap to Realize Memristive Three-Dimensional Neuromorphic Computing System

The Roadmap to Realize Memristive Three-Dimensional Neuromorphic Computing System

A Digital Liquid State Machine With Biologically Inspired Learning and Its Application to Speech Recognition

Step forward to map fully parallel energy efficient cortical columns on field programmable gate arrays

Contact Info

Product

Resources

About