8 October 2024 in IEEE Transactions on Circuits and Systems I: Regular Papers
This paper proposes an energy-and area-efficient sound localization neural network mimicking the auditory brainstem cognitive function. By adopting the bio-plausible Jeffress model, the proposed neural network locates the sound based on the interaural time difference (ITD) in an energy-and hardware-efficient manner. The proposed network modifies the original structure of the Jeffress model having a pair of long axon lines to provide performance gain. It can reduce power consumption and area by using a single axon line. It can further improve efficiency in terms of power and area by shortening the length of the axon line for pulse propagation. Since only the leading pulse is allowed to propagate through the shortened single axon delay line, the number of delay elements and corresponding network components are reduced. Moreover, it can accurately detect the location of the sound source thanks to the axon line composed of synchronized delay elements. A further reduction of the power consumption is achieved by eliminating redundant pulse propagation through the axon line after the output neuron fires. The proposed sound localization neural network was fabricated in a 28-nm CMOS process. The performance evaluation results indicate that the proposed sound localization neural network can detect the location of a sound source with a one-degree resolution at a given robot head size of 3.0125 cm, regardless of process corners. It also indicates that the network achieves up to 86.6% and 97.2% energy and area reduction from conventional sound localization networks, operating at 0.305-V supply voltage.
29 August 2024 in IEEE Transactions on Cognitive and Developmental Systems
In the sound localization of species, short-term depression (STD) plays an important role in maintaining interaural timing difference (ITD) sensitivity. In this paper, a biomimetic spiking neural network (SNN) utilizing 2D synaptic devices for mimicking biological sound localization is presented. A two-terminal monolayer device is used as the artificial synapse, whose temporal conductance change mimics the STD of a synapse. Alpha synaptic current and leaky integrate-and-fire (LIF) neuron models are used for realistic cortical operation. Lateral inhibition and superior olivary nucleus (SON) are adopted to increase the acuteness, to compensate for the interaural level difference (ILD)-induced disturbance, and to enlarge the sound intensity range. By combining solid-state STD synapses and bio-plausible cortical models with an ITD-based coincidence detection mechanism to mimic the auditory brainstem processing, our SNN achieved sound localization with a human-level resolution of one degree.