Scholarlink

<p> Background The freshness of fish is widely recognized as a key indicator of fish quality. Fish serves as a vital source of protein and essential nutrients for humans. To maximize the benefits of the proteins and nutrients found in fish, maintaining its freshness is critical. Most existing techniques are designed for whole fish and rely on visible external features such as the eyes, gills, and skin. This dependence limits the availability of non-invasive, real-world methods for detecting freshness. Additionally, there is a shortage of artificial intelligence models specifically designed to assess the freshness of packed fish. Objective This research aims to address these issues by developing an AI-powered E-nose system, named ‘FreshSense-SE’, for packed fish freshness classification using VOC sensors. This system accurately assesses fish freshness levels without physical sampling and chemical testing. <b>Methodology</b>: The proposed study begins with the collection of cut and packed fish samples stored under controlled conditions. The developed device employs several gas sensors, including MQ-135 (TMA), MQ-136 (H2S), MQ-137 (NH3), MLX90614 (temperature) to capture Volatile Organic Compounds (VOCs) emitted from the fish over time. The sensor signals are recorded and preprocessed to eliminate noise and normalize the data for efficient detection. The pre-processed data is then used to train the proposed detection model, which integrates a 1D-CNN with the squeeze-and-excitation attention model to detect fish freshness (i.e., fresh, semi-fresh, or spoiled). Finally, the trained model is deployed on a microcontroller for real-time freshness prediction. Result The experimental results showed that the accuracy of the proposed model in classifying fish freshness reached 99.39%. By integrating this AI system with a DL freshness classification model, we developed an efficient, non-invasive, real-time, and cost-effective fish freshness detection system that can be used to check the quality of the packed fish before cooking or directly consuming.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

FreshSense-SE: A Real-Time E-Nose system for packed fish freshness classification using VOC sensors

  • 期刊论文

  • 发表日期：

    2025年10月17日

查看全文
  • P. Kabitha,
  • D. Usha Nandini

摘要

Background The freshness of fish is widely recognized as a key indicator of fish quality. Fish serves as a vital source of protein and essential nutrients for humans. To maximize the benefits of the proteins and nutrients found in fish, maintaining its freshness is critical. Most existing techniques are designed for whole fish and rely on visible external features such as the eyes, gills, and skin. This dependence limits the availability of non-invasive, real-world methods for detecting freshness. Additionally, there is a shortage of artificial intelligence models specifically designed to assess the freshness of packed fish. Objective This research aims to address these issues by developing an AI-powered E-nose system, named ‘FreshSense-SE’, for packed fish freshness classification using VOC sensors. This system accurately assesses fish freshness levels without physical sampling and chemical testing. Methodology: The proposed study begins with the collection of cut and packed fish samples stored under controlled conditions. The developed device employs several gas sensors, including MQ-135 (TMA), MQ-136 (H2S), MQ-137 (NH3), MLX90614 (temperature) to capture Volatile Organic Compounds (VOCs) emitted from the fish over time. The sensor signals are recorded and preprocessed to eliminate noise and normalize the data for efficient detection. The pre-processed data is then used to train the proposed detection model, which integrates a 1D-CNN with the squeeze-and-excitation attention model to detect fish freshness (i.e., fresh, semi-fresh, or spoiled). Finally, the trained model is deployed on a microcontroller for real-time freshness prediction. Result The experimental results showed that the accuracy of the proposed model in classifying fish freshness reached 99.39%. By integrating this AI system with a DL freshness classification model, we developed an efficient, non-invasive, real-time, and cost-effective fish freshness detection system that can be used to check the quality of the packed fish before cooking or directly consuming.