Manuscript Number: APAC-D-24-02138  
手稿编号APAC-D-24-02138

Estimating the subjective severity of laptop fan abnormal sounds using psychoacoustic parameters
利用心理声学参数估计笔记本风扇异常声音的主观严重程度

Dear Dr Huang,
尊敬的黄博士

Thank you for submitting your manuscript to Applied Acoustics.
感谢您向《应用声学》投稿。

I have completed my evaluation of your manuscript. The reviewers recommend reconsideration of your manuscript following major revision. I invite you to resubmit your manuscript after addressing the comments below. Please resubmit your revised manuscript by Feb 06, 2025.
我已经完成了对您稿件的评审。审稿人建议对您的稿件进行重大修改后重新考虑。请您针对以下意见重新投稿。请在 2025 年 2 月 6 日之前重新提交您修改后的稿件。

When revising your manuscript, please consider all issues mentioned in the reviewers' comments carefully: please outline every change made in response to their comments and provide suitable rebuttals for any comments not addressed. Please note that your revised submission may need to be re-reviewed. 
在修改稿件时，请仔细考虑审稿人意见中提到的所有问题：请概述根据审稿人意见所做的每处修改，并对未处理的意见提出适当的反驳。请注意，您修改后的稿件可能需要重新评审。

To submit your revised manuscript, please log in as an author at https://www.editorialmanager.com/apac/, and navigate to the "Submissions Needing Revision" folder.  
要提交修改后的稿件，请以作者身份登录 https://www.editorialmanager.com/apac/，并导航至 "需要修改的稿件 "文件夹。

Research Elements (optional)
研究要素（可选）

This journal encourages you to share research objects - including your raw data, methods, protocols, software, hardware and more – which support your original research article in a Research Elements journal. Research Elements are open access, multidisciplinary, peer-reviewed journals which make the objects associated with your research more discoverable, trustworthy and promote replicability and reproducibility. As open access journals, there may be an Article Publishing Charge if your paper is accepted for publication. Find out more about the Research Elements journals at https://www.elsevier.com/authors/tools-and-resources/research-elements-journals?dgcid=ec_em_research_elements_email.
本期刊鼓励您分享研究对象，包括原始数据、方法、协议、软件、硬件等，这些对象支持您在《研究要素》期刊上发表原创研究文章。研究要素》是开放存取、多学科、同行评审的期刊，它使与您的研究相关的对象更容易被发现、更值得信赖，并提高了可复制性和可再现性。作为开放存取期刊，如果您的论文被接受发表，可能需要支付文章发表费。有关研究要素期刊的更多信息，请访问 https://www.elsevier.com/authors/tools-and-resources/research-elements-journals?dgcid=ec_em_research_elements_email。

Applied Acoustics values your contribution and I look forward to receiving your revised manuscript.
应用声学》重视您的贡献，我期待收到您修改后的稿件。

"For the latest news and updates, be sure to follow AA on Twitter @AcousticalSci!"
"如需了解最新消息和更新，请务必在 Twitter @AcousticalSci 上关注 AA！"

Kind regards,  
谨致问候、

Arianna Astolfi
阿里安娜-阿斯托尔菲  

Associate Editor  
副主编

Applied Acoustics
应用声学

Editor and Reviewer comments:
编辑和审稿人意见：

Reviewer #1: This paper presents a classical study of the acoustic quality of a product (laptops here). However, some methodological choices are difficult to understand and require further explanation - they are listed below :
审稿人 1：本文介绍了一项关于产品（此处为笔记本电脑）声学质量的经典研究。但是，有些方法选择难以理解，需要进一步解释--下面列出了：

- Stimuli are presented by a loudspeaker. Though the paper says that "each participant is seated facing the loudspeaker" (2.1.3 line 46) , fig.3 shows this is not the case. It seems that the directivity characteristics (or, in more general words, the frequency response in that direction) of the loudspeaker has not been corrected. Is this correct ? In that case, I recommend to recorder the stimuli (emitted by the loudspeaker) before conducting the computation of sound metrics.
- 通过扬声器呈现刺激。虽然论文中说 "每个参与者都面向扬声器而坐"（2.1.3 第 46 行），但 图 3 显示情况并非如此。似乎扬声器的指向性特性（或用更通俗的话说，该方向上的频率响应）没有得到修正。这是否正确？在这种情况下，我建议在计算声音度量之前记录（扬声器发出的）刺激。

- Roughness values are quite low (below 0.01 asper in Exp.1 and 0.06 asper in Exp.2 : to be compared to the values cited in Annex C of the ECMA-418-2 standard, which are between 0.2 and 1.2). Therefore, I do not think that roughness can be considered as a good predictor, even though statistics suggest otherwise.
- 粗糙度值相当低（Exp.1 中低于 0.01 asper，Exp.2 中低于 0.06 asper： 与 ECMA-418-2 标准附件 C 中引用的值相比，后者介于 0.2 和 1.2 之间）。因此，我认为粗糙度不能被视为一个很好的预测指标，尽管统计数据表明并非如此。

- Was the consensus between participants evaluated ? As loudness is the main (only) contributor, a high consensus can be expected - but it would be useful to check it out. Or is this represented by Cronbach' alpha, mentioned in 2.3.2 (line 60) ? I do not understand how this number was computed (basically, Cronbach's alpha is used to evaluate the consistency of a questionnaire made of several questions, which is clearly not the case in this study).
- 是否评估了参与者之间的共识？ 由于响度是主要的（唯一的）影响因素，因此可以预期会有很高的共识--但最好还是核实一下。 我不明白这个数字是如何计算出来的（基本上，Cronbach's alpha 是用来评估由多个问题组成的问卷的一致性的，而本研究显然不是这种情况）。

- In Exp.2, loudness was equalized within each group of equal fan rpm. Why not along the whole set of sounds ? A major limitation of the choice made by authors is that sharpness is again correlated with loudness (the hidden variable being the fan rotational speed).
- 在实验 2 中，每组风扇转速相同时，响度是均衡的。为什么不在整组声音中？ 作者所做选择的一个主要局限是，尖锐度再次与响度相关（隐藏变量是风扇转速）。

- In Exp.2, stimuli loudness is much higher than in Exp.1 (between 4.5 and 11 sone GF vs. below 4 sone GF). This was done in order to "enhance participants' perception of the abnormal sound" (3.1.1 line 32) - at the expense of the ecological validity of the study. To my point of view, this reduces the applicability of the results.
- 在实验 2 中，刺激物的响度比实验 1 高得多（4.5 到 11 sone GF 之间，而不是低于 4 sone GF）。这样做是为了 "增强参与者对异常声音的感知"（3.1.1 第 32 行）--以牺牲研究的生态有效性为代价。在我看来，这降低了研究结果的适用性。

Some other remarks :
其他一些评论：

- It is difficult to admit that sound from laptops can induce psychological depression, as it is implied at the very beginning of the paper. Ref. 1 refers to overall noise in hotels, in which I guess the contribution from laptop is very low.
- 很难承认笔记本电脑发出的声音会引起心理压抑，因为本文一开始就暗示了这一点。参考文献参考文献 1 指的是酒店的整体噪音，我想其中来自笔记本电脑的噪音很小。

- Tonality is not limited to "a single-frequency component or narrow-band components", as written in 2.2.1 (line 17) : a sound with several harmonic components has a high tonality, according to ECMA 418-2.
- 音调并不局限于 2.2.1（第 17 行）中所写的 "单一频率成分或窄带成分"：根据 ECMA 418-2，具有多个谐波成分的声音具有高音调。

- In 2.3.2 (line 45), authors said that "non-parametric testing methods are appropriate for subsequent analyses". But, in the following, it seems to me that parametric methods have been used - or some information may be missing (as an example, how was computed the correlation values in tables 1 and 2 ?).
- 在 2.3.2（第 45 行）中，作者说 "非参数测试方法适合后续分析"。但在下文中，我觉得似乎使用了参数方法--或者可能缺少某些信息（例如，表 1 和表 2 中的相关值是如何计算的？

Reviewer #2: Review of Manuscript Draft:
审稿人 2：审稿：

Estimating the subjective severity of laptop fan abnormal sounds using psychoacoustic parameters
利用心理声学参数估算笔记本风扇异常声音的主观严重程度

This study investigates the severity of abnormal laptop fan noise, which, in recent years, become one of the prominent sources of annoyance due to the IT-equipment noise as well as household appliance noises. Study delivers very important results for the manufacturers and the researchers on the relationship between abnormal laptop fan noise and the psychoacoustic metrics.
本研究调查了笔记本电脑风扇异常噪音的严重程度，近年来，这种噪音已成为 IT 设备噪音和家用电器噪音的主要烦恼源之一。该研究就笔记本电脑风扇异常噪音与心理声学指标之间的关系为制造商和研究人员提供了非常重要的结果。

In two experiments, the correlations between perceived severity of the fan noise examples and the psychoacoustic metrics are thoroughly investigated. At the end, a severity evaluation model was built highlighting the important correlation between loudness and roughness and estimated severity. The developed model performs quite well in the defined constraints.
在两个实验中，对风扇噪声示例的感知严重程度与心理声学指标之间的相关性进行了深入研究。最后，建立了一个严重性评估模型，突出了响度和粗糙度与估计严重性之间的重要相关性。所开发的模型在规定的限制条件下表现良好。

The extensive number of laptops recorded, the recordings obtained under two different scenarios, and the systematic collection of recordings at varying RPMs demonstrate that this study has a highly detailed input set. In this regard, the comprehensiveness of the input set indicates a high level of general validity for the findings. The listening tests conducted with a dataset of high representativeness are also robust and well-structured. The statistical analyses that support the study's conclusions are sufficient.
记录的笔记本电脑数量之多、在两种不同情况下获得的记录，以及在不同转速下系统收集的记录，都表明这项研究具有高度详细的输入集。在这方面，输入集的全面性表明研究结果具有很高的普遍有效性。使用具有高度代表性的数据集进行的听力测试也是稳健且结构合理的。支持研究结论的统计分析也很充分。

To enhance the readability and effectiveness of the study, I propose the following minor changes:
为提高研究报告的可读性和有效性，我建议做以下小改动：

Chapter 2.1.1.-Stimuli
第 2.1.1.章--刺激

Line 43: What is the "elevated use" scenario was thought for? For any possible dockstation application etc? Maybe it would be better to describe this scenario in text, a little bit in more detail. (maybe even with additional figure?)
第 43 行："高架使用 "的设想是什么？用于任何可能的对接站应用等？也许用文字更详细地描述一下这个方案会更好。(也许甚至可以加上额外的数字？)?

Line 7-9 (next page): More detailed information on selection of 72 stimuli could be influential. When the expert panel was listening, which aspects were they focusing on? At the end, how many laptops (amongst 18) were included in 72 stimuli?
第 7-9 行（下一页）：关于 72 种刺激物的选择的更多详细信息可能会有影响。专家小组在听取意见时，重点关注了哪些方面？最后，72 个刺激点中包含了多少台笔记本电脑（在 18 个刺激点中）？

Line 15-21 & Figure 2: That would be easier to follow, if a figure could be added for "the same laptop with different RPMs", where all the lines are in one graph. Similarly "different laptops in the same RPM" could be also an interesting figure, to explain the variance amongst different laptops. These could be either an FFT where all the lines are together in different colors, or a spectrogram, where the stimuli were put side by side, so that the comparison is easier.
第 15-21 行 & 图 2：如果能为 "转速不同的同一台笔记本电脑 "添加一个图，让所有线条都在一个图中，那就更容易理解了。同样，"相同转速下的不同笔记本电脑 "也可以是一个有趣的数字，以解释不同笔记本电脑之间的差异。这可以是一个 FFT 图，其中所有线条都以不同的颜色显示在一起，或者是一个频谱图，将刺激并排显示，这样就更容易进行比较。

Line 15-21 & Figure 2: Do the sub-figures (a) (b) and (c) originate from the same laptop?
第 15-21 行和图 2：子图 (a) (b) 和 (c) 是否来自同一台笔记本电脑？

Line 15-21 & Figure 2: I am aware of the fact that we could not describe what subjective severity stands for, before the actual results of the listening tests, but a little bit of speculative description of what could be considered as "sever" would be essential at this point. In that way, reader can follow the 'idea of severity' with a possible electro-mechanical reasoning in mind. For example, does severity differences between Fig2 d-e-f was created because of the differences around 100 Hz, tonality around 5 kHz, or general broadband level increase around 1kHz? Because, the 1 kHz levels of e and f seem to be similar, but the tonality around 5 kHz not.
第 15-21 行和图 2：我知道，在实际听音测试结果出来之前，我们无法描述主观严重性的含义，但在这一点上，对什么是 "严重 "进行一点推测性描述是必不可少的。这样，读者就可以根据 "严重程度的概念 "来推理可能的电子机械原理。例如，Fig2 d-e-f 之间的严重程度差异是由于 100 赫兹左右的差异、5 千赫左右的音调差异，还是由于 1 千赫左右宽带电平的普遍提高？因为，e 和 f 的 1 千赫兹电平似乎相似，但 5 千赫兹附近的音调却不同。

Lines 54-55: Two comments on that:
第 54-55 行：有两点意见：

1- Maybe it is better to give also the dB(A) values of the selected stimuli. They could be better to grasp for every reader than loudness values.
1- 也许最好同时给出所选刺激的 dB(A) 值。对于每个读者来说，这些值可能比响度值更容易掌握。

2- What about the background noise of the listening test facility? Maybe it would be better to give in a separate figure (FFT avg) how the stimuli lie above the background noise.
2- 听力测试设施的背景噪声如何？也许最好用单独的图表（FFT 平均值）说明刺激物如何高于背景噪声。

Chapter 2.2.1-Psychoacoustic Metrics
第 2.2.1 章--心理声学指标

Lines 11-12: DIN 45692 calculations was based on Aures model, which incorporates already the effect of loudness of the stimuli. Have you ever investigate the effect and correlation of the Bismarck sharpness model? What I understand from FFTs, that the stimuli can differentiate in the higher frequency content (Figure 2 e and f), even for similar levels.
第 11-12 行：DIN 45692 标准的计算是基于Aures模型的，该模型已经包含了刺激响度的影响。您是否研究过俾斯麦锐度模型的效果和相关性？我从 FFT 中了解到，刺激物可以在较高频率的内容中产生差异（图 2 e 和 f），即使在相似的电平下也是如此。

Chapter 2.3.1. Psychoacoustic Metrics Distribution
第 2.3.1 章.心理声学指标分布

Line 50: Is it possible to explain the underlying mechanism why the roughness was decreased by increasing RPM?
第 50 行：是否有可能解释为什么通过提高转速来降低粗糙度的内在机制？

Line 54: It is interesting to see that the hearing model tonality was delivering N/A tonality values. What I can imagine, for example, from Figure 2. (f) that the tones around 5 kHz would be audible, considering the absolute level of the tone and the level differences between tone and noise. (Although, this consideration is purely subjective).
第 54 行：有趣的是，听力模型音调提供的是不确定音调值。例如，从图 2（f）中我可以想象，考虑到音调的绝对电平以及音调与噪声之间的电平差异，5 kHz 左右的音调是可以听到的。(尽管这种考虑纯属主观）。

In general, as a general engineering intuition, I would not consider the laptop fan noise is being "not tonal", from this aspect, that was somewhat unexpected. That would be really interesting to calculate/catch the tones using another method (TNR, PR) and report that. Even for correlations, the frequency range could be divided into couple of segments (maybe for low frequencies 100 Hz etc. and high frequencies for any tones above 4 kHz etc.) and calculate the tonality for these regions as single values. These single values can be used for correlations. There might be examples in the literature for that approach, would be interesting to see if the same approach delivers different results.
总的来说，作为一般的工程直觉，我不会认为笔记本电脑风扇的噪音 "没有音调"，从这个角度来看，这有点出乎意料。如果能用另一种方法（TNR、PR）计算/捕捉音调并报告，那将会非常有趣。即使是在相关性方面，也可以将频率范围分为几个部分（也许低频为 100 Hz 等，高频为 4 kHz 以上的任何音调等），并将这些区域的音调性计算为单一值。这些单一值可用于相关性分析。文献中可能有这种方法的例子，看看同样的方法是否会产生不同的结果会很有趣。

Chapter 2.3.3-Severity Model
第 2.3.3 章-严重程度模型

Line 20-21, Table 1: I may be wrong, but is it possible to add the RPM as a parameter in these analyses? Eventually, what I can imagine, that the relationship between severity perception and calculated parameters could be different for low and high rotating speeds. When calculating these correlations in this way, aren't we effectively averaging the effects across all RPMs?
表 1 第 20-21 行：我可能弄错了，但是否有可能在这些分析中加入转速作为参数？最终，我能想到的是，低转速和高转速的严重性感知与计算参数之间的关系可能不同。在以这种方式计算这些相关性时，我们不是有效地平均了所有转速的影响吗？

Line 1, Table 3 (next page): How much R2 improvement was gained by adding the roughness into the equation? That would be a fine way of formulating the effectiveness of the developed model.
表 3 第 1 行（下一页）：将粗糙度加入等式后，R2 提高了多少？这将是表述所开发模型有效性的一个很好的方法。

Chapter 3.1.1.-Stimuli
第 3.1.1.1 章--刺激

Lines 25-26: In the experiment 2, same RPM steps were used as in Exp 1. However, this time, the obtained loudness values were way higher than experiment 1, according to the Table 4. (Figure 5 says, for example, the loudness values of 2700 RPM were around 1 soneGF and below, but in experiment 2, the values have mean value of 4.65 soneGF). What could be the reason for this difference?
第 25-26 行：在实验 2 中，使用了与实验 1 相同的转速步骤。然而，根据表 4，这次获得的响度值远远高于实验 1。(例如，图 5 显示，2700 RPM 的响度值约为 1 soneGF 及以下，但在 实验 2 中，该值的平均值为 4.65 soneGF）。造成这种差异的原因是什么？

Lines 29-30: Similar to the Experiment 1, it would be better to explain; from how many different laptops these 64 samples were selected.
第 29-30 行：与实验 1 类似，最好解释一下这 64 个样本是从多少台不同的笔记本电脑中挑选出来的。

Line 45: Similar to the Experiment 1: More information about the possible sources/reasons of severity would be reasonable.
第 45 行：与实验 1 类似：最好能提供更多有关严重性可能来源/原因的信息。

Chapter 3.1.3. Data Analysis
第 3.1.3 章.数据分析

Lines 46-48: I guess similar to my question for the first listening experiment: loudness equalization has been done within the RPM steps. Still, when considering all the RPM steps together in the experiment, there is a loudness difference between stimuli. (Figure 10- Loudness is still an input with variance). Then, analysis of the results might be done separately for each RPM? Or related to Table 5, between 64 stimuli, there are loudness differences. If all the stimuli were asked consecutively in an experiment, then we might need to consider again the effect of loudness? According to Figure 10, loudness can still explain some of the variance in severity rating score in Experiment 2, however, expectedly lower than in Experiment 1.
第 46-48 行：我想这与我在第一个听音实验中提出的问题类似：响度均衡已在 RPM 步骤中完成。尽管如此，在实验中将所有 RPM 步骤放在一起考虑时，刺激物之间仍存在响度差异。(图 10--响度仍然是一个有差异的输入）。那么，是否可以对每个转速分别进行结果分析？或者与表 5 有关，64 个刺激之间存在响度差异。如果在实验中连续提出所有刺激，那么我们可能需要再次考虑响度的影响？根据图 10，响度仍然可以解释实验 2 中严重程度评分的部分差异，但预计会低于实验 1。

Chapter 4.2.-Validation of the Subjective Severity Evaluation Model
第 4.2 章--主观严重性评价模型的验证

Lines 1-2 (next page): Here the statistical analyses based on RPM were given. As I have mentioned in the Experiment 1 and 2, I find these considerations quite useful, and I guess it would make the paper even better to include such investigations in the corresponding chapters for Exp 1 and 2.
第 1-2 行（下一页）：这里给出了基于 RPM 的统计分析。正如我在实验 1 和 2 中提到的，我认为这些考虑因素非常有用，我想如果能将这些调查纳入实验 1 和 2 的相应章节中，会使论文更加完善。

Chapter 5. Conclusion
第 5 章.结论

As I explained in my former comments, I think it would be quite useful and reasonable to mention any possible physical reasoning behind the severity perception. It seems reasonable to mention, even briefly, the implications of the findings for device design. Which design feature makes a laptop fan to emit "less severe" noise? What kind of noise was found to be abnormal and what was the underlying mechanism behind? I can imagine that some of the information is confidential, but at least some brief information would make the paper way stronger, in my opinion.
正如我在之前的评论中所解释的，我认为提及严重性感知背后任何可能的物理原因都是非常有用和合理的。即使是简短地提及研究结果对设备设计的影响，似乎也是合理的。哪种设计特点使笔记本电脑风扇发出的噪音 "不那么严重"？哪种噪音被认为是不正常的，其背后的机制是什么？我可以想象有些信息是保密的，但在我看来，至少一些简短的信息会让论文更有说服力。

At the end, this paper shows a very profound scientific method, well-defined hypotheses, proper experimentation and analyses. With briefly addressing the comments above, this paper is going to be highly beneficial to researchers working in this field.
最后，这篇论文展示了非常深刻的科学方法、明确的假设、适当的实验和分析。通过对上述意见的简要回应，这篇论文将对这一领域的研究人员大有裨益。