Memristor chip cuts energy and speeds up depression detection
Energy-Efficient Memristor Accelerator for Depression Detection with a CNN-LSTM Hybrid Model
Hardware Architecture
Summary
Detecting depression early using speech data can take a lot of computing power, which wastes energy and time. The authors designed a special computer chip that uses tiny resistors called memristors to run a combined CNN-LSTM model for analyzing speech more efficiently. This chip does many steps together in one place, so it moves less data around and runs much faster than regular CPUs or GPUs. The new design reduces energy use by over 100 times and speeds up processing by around 30 times, though prediction accuracy still needs checking.
What this means in practice
- •For embedded system engineers: Build low-power hardware that quickly analyzes speech patterns for early depression screening in wearable or mobile devices.
- •For audio processing hardware designers: Design specialized accelerators that combine CNN and LSTM models on memristor crossbars to reduce latency and energy consumption in audio analysis.
Tested on simulated data.
Authors
Jiaqi Liu, Yun Yu, Fenglin Xi, Shenjian Zhang, Shenghao Liao, Wenze Zheng, Mingxuan Hu, Yangbin Chen, Xinhong Chen, Chun Zhao
Abstract
Early screening of depression requires efficient analysis of audio and speech data, while conventional CPU and GPU platforms incur high latency and energy consumption for CNN-LSTM inference. To address this challenge, we propose a unified resistive RAM (ReRAM) processing-in-memory accelerator for a CNN-LSTM attention workload. The architecture integrates convolutional, recurrent, attention, and multi-layer perceptron (MLP) stages within a crossbar-based framework, reducing intermediate data movement between spatial and temporal branches. CNN operations are mapped through im2col-based vectorization, while LSTM gate computations are reformulated using concatenated input-state vectors and mapped onto ReRAM crossbars. We further provide module-level workload breakdown, parameter sensitivity analysis, and matched-workload context against representative ReRAM/PIM accelerators. Experimental results show 32.1 times and 9.3 times latency reduction over CPU and GPU baselines, respectively, with about 160 times and 140 times energy reduction. Simplified weight-perturbation tests characterize ASLF sensitivity in an untrained pipeline with synthetic spectrogram-shaped inputs. Task-level prediction accuracy and robustness remain to be evaluated.