| IP명 | Bio-inspired Energy-efficient Sound Localization Neural Network based on Auditory Brainstem Processing | ||
|---|---|---|---|
| Category | Analog | Application | AI |
| 실설계면적 | 4㎛ X 4㎛ | 공급 전압 | 1V |
| IP유형 | Hard IP | 동작속도 | 100MHz |
| 검증단계 | Silicon | 참여공정 | SS28-2301 |
| IP개요 | We proposes an energy-efficient bio-inspired sound localization neural network mimicking the auditory brainstem cognitive function. The proposed neural network can perform the localization of a sound source based on the interaural time difference (ITD) information from both ears in a very energy-efficient manner. The proposed neural network is composed of CMOS synapses, neurons, axon delay lines, and other components, mimicking the auditory brainstem neural structure. The proposed sound localization neural network will be designed in a 28-nm CMOS process. The pre-layout performance evaluation results indicate that the network achieves at least 82.7% power reduction as compared to conventional sound localization networks. Contributed by a processinvariant operation, the proposed neural network supports onedegree sound localization resolution with 100% accuracy at various process corners with 0.3 V supply voltage. |
||
- 레이아웃 사진 -
|
|||