Precise Identification of Food Smells to Enable Human–Computer Interface for Digital Smells

Round 1

Reviewer 1 Report

The paper describes an approach using a convolutional neural network (1D-CNN) algorithms to classify the sensor responses to VOCs emitted by food compounds.

The paper is well structured and referenced. The methods are clearly described and the results clearly presented. The technical The 1D-CNN approach appears to show an advantage in terms of classification accuracy compared to other machine learning.

Line 37: What is meant by food designators? A brief explanation would be helpful.

Line 95: Can the authors state the manufacturer of the MOX sensors?

Line 251: “This is because 1D convolutional layers can capture better features than exponential fitting” Can the authors clarify what is meant by this statement in relation to the sensor characteristics.

The characteristics of MOX sensors are known to drift over time and drift can create false features. Can the authors comment on this and what steps were taken to minimise / mitigate the effects of MOX sensor drift?

Author Response

in the attached response (pdf) document

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors demonstrated the use of 1D-CNN for identification of fruits and their decomposition monitoring.

 

First of all, the authors may directly transfer the manuscript from something IEEE paper to Electronics without editing. This kind of manuscript is not suitable for peer-reviewing at all.

 

Line 64: To the best of our knowledge...
There are several reports on such topics, e.g., 10.1149/1945-7111/ac1699.
Comparing those reports, the authors did not clearly describe the differences or improvements.

Sample size: The authors described each fruit smell contains 15 samples in line 123. Did the authors measure 15 individual fruits or 1 fruit with 15 times measurements?
If the former case, the authors should show all signal responses.
If the latter case, the authors should address the reproducibility.

 

Decomposition monitoring: the authors measured 20 samples for each storage day. Did the authors measure 20 individual fruits for each fruit for each storage day? If so, the authors should address the reproducibility.
Also, did the authors measure the series of 4-days measurement in different date? If the author performed single measurement, it is not clear that the authors monitored the decomposition or the effects of others, such as humidity, temperature, and so on. The authors should carefully measure these points.

 

PCA: The authors should show PCA plots.

 

Author Response

in the attached response to reviewer 2 document (pdf)

Author Response File: Author Response.pdf

Reviewer 3 Report

There are a few errors of the language type, which can be easily corrected.

Line 97: detected scents are not „digital” but will be digitized.

Lines 105,106: expression „different exposure” is repeated 2 times.

In Figures 2 and 6 you use term „avacado” instead of „avocado”.

Line 301: there is a grammatical error in the last sentence of Conclusion.

Additional remark concerns the sensor recovery. From the text as in lines 174, 175 it is not clear where this recovery is shown. If it concerns Figure 2, it should be precisely indicated in the figure.

Author Response

in the attached response to reviewer 3 document (pdf)

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The authors addressed the reproducibility while the results are not in the main text.

Reproducibility measurements for decomposition monitoring: Although the authors attempted repeated measurements over a four-month period, the results do not indicate that the model in this study can identify fruit decomposition by smell, independent of external factors such as daily humidity and temperature. The authors stated in the response that the humidity and temperature were in the range of 23–27 °C and 40%–60%. The reviewer believes that the range of temperature and humidity is large. The optimized 1D-CNN model may extract the features which relate the difference of the temperature and humidity instead of the differences in smells.

If the authors claim that "This work provides a valuable reference to digitize scents for precise food identification and decomposition monitoring," the authors should demonstrate the identification of 5 different fruit by using the model built on different samples, or identification of 5 different fruit using at least three individual fruit for each category. Also, the authors should clearly show the decomposition monitoring results are independent of the external effects, such as humidity and temperature.

Overall, the reviewer believes that the research design is not consistent with the conclusion claimed by the authors.

Author Response

in the attached document

Author Response File: Author Response.pdf