Journal of Semiconductors, Volume. 46, Issue 1, 012604(2025)

Flexible artificial vision computing system based on FeOx optomemristor for speech recognition

Jie Li1, Yue Xin1, Bai Sun2、*, Dengshun Gu1, Changrong Liao3, Xiaofang Hu1, Lidan Wang1, Shukai Duan1, and Guangdong Zhou1、**
Author Affiliations
  • 1College of Artificial Intelligence, Southwest University, Chongqing 400715, China
  • 2Frontier Institute of Science and Technology, Xi’an Jiaotong University, Xi’an 710049, China
  • 3School of Electronic Information and Electrical Engineering, Chongqing University of Arts and Sciences, Chongqing 402160, China
  • show less
    Figures & Tables(4)
    (Color online) Flexible optomemristor based on C27H30O15/FeOx heterostructure. (a) Purple lotus solution prepared from lotus flowers, with the molecular structure of C27H30O15 lotus in the switch functional layer of the optoelectronic memristor below. (b) Diagram illustrating the simulation of synapses and device structure of the Au/C27H30O15/FeOx/ITO memristor. (c) I−V characteristic curves of the Au/C27H30O15/FeOx/ITO optoelectronic memristor under light and dark conditions, showing a significant positive light effect with increased current density under light exposure. (d) Optical bandgap energy fitting comparison of FeOx and C27H30O15/FeOx UV−Vis (Ultraviolet–visible) spectra. (e) Double logarithmic SCLC mechanism fitting under HRS state of the device. (f) Calculation of defect concentration at different I − V scan cycles (iterations) under HRS state. (g) Diagram illustrating the Au/C27H30O15/FeOx/ITO flexible memristor at different curvature radii.
    (Color online) Electrical characteristics of Au/C27H30O15/FeOx/ITO. (a) Resistance modulation of Au/C27H30O15/FeOx/ITO memristor under different bias scan voltages. (b) Modulation of memory behavior with different bias voltage scan rates. (c) I−V characteristic curves of C27H30O15/FeOx memristor tested after bending at a 60° angle for 1, 101, 102, 103, 104 times, demonstrating excellent resilience of the flexible device to various bends. (d) Cycle endurance test of the memristor, with Au/C27H30O15/FeOx/ITO memristor achieving at least 103 scan cycles. (e) Duration of high resistance state (HRS) and low resistance state (LRS) at Vread = 0.3 V. (f) 32 non-volatile multi-conductive states lasting 300 s, capable of achieving at least 5-bit computational accuracy. (g) Device-to-device stability. (h) Double logarithmic SCLC mechanism fitting in device LRS state. (i) Calculation of defect density at LRS state under different I − V scan cycles (iteration counts).
    (Color online) Optical Characteristics of Au/C27H30O15/FeOx/ITO. (a) Under light conditions, the cycle endurance of the Au/C27H30O15/FeOx/ITO memristor reaches up to 103 cycles. (b) Light pulse intensity modulates the photoconductive state of the C27H30O15/FeOx memristor. (c) Light pulse width modulates the photoconductive state of the C27H30O15/FeOxmemristor. (d) At different light intensities, the photocurrent shows a linear correlation with frequency. (e) and (f) Gradual linear adjustment of 150 photoconductive states has the potential to achieve a calculation accuracy of 7 bits. Use a light pulse (19 mW, 0.5 s) (upper part of Fig. 3(f)) to stimulate the device, 150 photoconductive states were obtained, as shown in Fig. 3(e). In order to further observe the 150 stable states of photoconductivity, the photoconductivity stabilized by applying light pulses to (e) was amplified to obtain (f). From (f), it can be seen that the photoconductivity state of the flexible device can be maintained for at least 16 s, indicating stability. (g) Under light pulses (19 mW, 0.5 s), the analog memristor exhibits a time-dependent photocurrent response. Adjusting the number of light pulses enables transitions from STP to LTP. (h) Characteristics of paired-pulse facilitation (PPF) in the Au/C27H30O15/FeOx/ITO memristor under light pulses (64.6 mW) at intervals (0.5−13 s) . (i) Electro−optical synergy conductivity update. The conductivity is increased from low to high using ten light pulses (19 mW, 0.5 s), and decreased from high to low using four negative voltage pulses (−0.1 V, 0.5 s).
    (Color online) Digital speech signal recognition system based on flexible optoelectronic memristors Au/C27H30O15/FeOx/ITO. (a) Schematic of a parallel RC system based on dynamic memristors. Each memristor RC unit has a different mask sequence. The output is a linear combination of all reservoir states. This parallel RC system is achieved by testing individual memristors over multiple cycles. The output vector is a linear combination of values from virtual nodes, and weights (Wout) can be trained using linear regression. (b) 16 encoding state diagrams of memristors, reflecting previous temporal information through different photocurrents. (c) Prediction results from the parallel RC system based on memristors. The purple line represents actual outputs, while the pink line represents predicted outputs of the RC system. The mask length is 9, with 25 masks. The training set: test set ratio is 7 : 3. (d) Confusion matrix of prediction results versus correct outputs from the RC system based on Au/C27H30O15/FeOx/ITO memristors, achieving an accuracy of 94.88%. The color bar indicates the normalized probability of each predicted result under the correct output.
    Tools

    Get Citation

    Copy Citation Text

    Jie Li, Yue Xin, Bai Sun, Dengshun Gu, Changrong Liao, Xiaofang Hu, Lidan Wang, Shukai Duan, Guangdong Zhou. Flexible artificial vision computing system based on FeOx optomemristor for speech recognition[J]. Journal of Semiconductors, 2025, 46(1): 012604

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Research Articles

    Received: Aug. 2, 2024

    Accepted: --

    Published Online: Mar. 6, 2025

    The Author Email: Sun Bai (BSun), Zhou Guangdong (GDZhou)

    DOI:10.1088/1674-4926/24080004

    Topics