空氣傳遞超音波模組整合於自由移動物體追蹤與即時聲波照射系統

Abstract

傳統超音波治療多仰賴麻醉或機械固定以限制動物活動，確保聲場聚焦之穩定性與治療準確度，然而此操作方式不僅降低實驗靈活性，亦對動物福祉造成負擔，並限制其於長期監測與重複性刺激等應用情境之可行性。為突破此限制，本研究提出一套整合超音波感測、機器學習預測與相位陣列控制之即時聚焦系統，實現對自由移動目標之非接觸式追蹤與能量聚焦。系統以超音波接收訊號為輸入特徵，透過 PointNet 模型即時預測目標頭部座標，進一步驅動 10×10 超音波相位陣列進行聚焦控制。整體架構可於無影像輔助情境下獨立運作，具備即時追蹤與對應聚焦能力。實驗結果顯示，系統可穩定產生對應位置之聚焦聲壓，量測結果與理論預期高度一致，展現良好之準確性與穩定性。另透過週期性訊號調變與脈波寬度調變，可實現 40 Hz 或更低頻率之刺激模式，進一步拓展其於神經調控、節律誘發等應用場域之潛力。

Conventional ultrasound therapies often rely on anesthesia or mechanical restraints to limit animal movement, ensuring stability of the acoustic focus and treatment accuracy. However, such procedures reduce experimental flexibility, impose burdens on animal welfare, and constrain the feasibility of long-term monitoring and repeated stimulation applications. In response to these limitations, this study develops a real-time focusing system that integrates ultrasonic sensing, machine learning–based prediction, and phased array control, enabling non-contact tracking and energy focusing on freely moving targets. The system utilizes received ultrasonic signals as input features and employs a PointNet-based model to predict the target’s head position in real time, which subsequently drives a 10×10 ultrasonic phased array for focusing control. The overall architecture can operate independently without visual assistance, supporting real-time tracking and corresponding focal control. Experimental results demonstrate that the system consistently generates focused acoustic pressure at the predicted target positions, with measured values closely matching theoretical expectations, indicating strong accuracy and stability. Furthermore, through periodic signal modulation and pulse width modulation, the system can deliver stimulation patterns at 40 Hz or lower, expanding its potential for applications in neural modulation, rhythm entrainment, and other behavioral research contexts.