基於聲音辨識實作智慧垃圾桶

Intelligent Garbage-Can based on Sound Recognition

李承峰 ； 林良翰 ； 賴文彬 ； 邱楷程 ； 周立德

《TANET2019 臺灣網際網路研討會》 2019年 (2019/12) Pp. 441-445

https://doi.org/10.6924/TANET.201909.0079

摘要

現今已有利用影像辨識技術進行分類的智慧垃圾桶，但由於垃圾的表面材質繁多是影像辨識的一大困難。本論文實作出以聲音來辨識垃圾種類的智慧垃圾桶，依據垃圾種類掉落撞擊所發出不同的聲響，使用麥克風收音，將樹莓派當作控制中心，輔以伺服馬達、超聲波感測器等硬體設備，以及建立在Keras機器學習套件的Convolutional Neural Networks（CNN）模型完成分類，並且利用雲端即時資料庫Firebase紀錄當前垃圾數量，在垃圾達到定量時提醒使用者。

關鍵字

深度學習 ； 聲音辨識 ； 卷積神經網絡

並列摘要

There has been some implementation using image recognition to do garbage classification, but the material of garbage is a major difficulty in image recognition. This research is a smart trash can identifying the type of garbage based on the sound made by the garbage. The sound is recorded by the microphone, and the Raspberry Pi is used as a control center, supplemented by hardware devices such as servo motors and ultrasonic sensors. Lastly, the Convolutional Neural Networks (CNN) model, built-in the Keras machine learning kit, is used to complete the classification. Moreover, a cloud database, Firebase, is used to record the current amount of garbage, and it alerts users when the amount of garbage reaches the threshold.

並列關鍵字

Deep Learning ； Sound Recognition ； Convolutional Neural Networks