Bioinspired Auditory Model for Vowel Recognition
Round 1
Reviewer 1 Report
The paper focuses on proposing a vowel recognition approach using an artificial neural network. Given the fact that the topic of voice recognition has been extensively studied over the last decades, the authors should better highlight which are the advantages of the proposed approach. Using a multilayer neural network can be considered a fairly simple approach.
# Additional comments
In section 2 the author mention the concept of "inverse labial radiation model block". The concept should be better explained for readers less familiar with the subject. The concept should also be probably put into connection with the figure above.
Also in section 2, the paper mentions the concept of lateral inhibition block. The concept should be better explained for readers less familiar with the subject. The concept should also be probably put into connection with the figure above.
Further details should be provided concerning the conditions for the audio recordings in order to make the presentation technically sound. For example, what was the level of the background noise and how did it relate to the level of the voice?
The authors are kindly asked to specify why they have chosen to reduce the quality of the recordings. While the technique is comun, especially in the case of video analysis, it would be better to highlight the reasons.
The labeled recordings should be shared using a public repository (github / zenodo) to promote research reproducibility.
The paper is missing comparisons with other works in the scientific literature concerning voice recognition. The authors are kindly asked to highlight which are the benefits of the proposed approach.
Author Response
Dear Reviewer,
Thank you for your insightful suggestions that help improve the understanding and scope of our work.
The article focuses on proposing a bio-inspired hearing model that allows vowel recognition. In this, an artificial neural network (ANN) is used as part of the human auditory system model, due to its similarity with biological systems. The use of a multilayer ANN can be considered not complex, but precisely because of its similarity to biological systems it is what makes it interesting to be part of the proposed model. At the same time, this proposal constitutes another approach to vowel recognition based on bio-inspired hearing models. Its scope or future importance is that it can be used to assessment personalized cochlear implant strategies before to be implanted, although it can also be used in the field of robotics and pattern recognition.
Their correct suggestions were taken into account in the corrections made.
Kind regards,
The authors.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This paper presents a bioinspired or neuromorphic model to replicate the vowel recognition process for the auditory system. The authors implement a bioinspired peripheral and central auditory system model, and propose a neuromorphic higher auditory system model based on artificial neuronal nets for vowel recognition. They conclude that the spectral representation of the output of the human phonation model corresponds to the spectral characteristics of the output of the cochlea, so that the modeling of the peripheral auditory system is allowed.
The paper has a good potential for being appreciated and cited, but it requires some important improvements and also extensions.
Comments:
- Introduction should specify more information with regard to the problem definition and scope of the paper.
- Highlight in what measure and in what parameters, the proposed methodology was found better as compared to existing ones, in the Introduction.
- A subsection/paragraph related to the problem analyzed should be included. The connection between the problem and the solution proposed is also not pointed out.
- Section related work has to be added. And each paper should clearly specify what is the proposed methodology, novelty and results cum experimentation. At the end of related works, highlight in some lines what overall technical gaps are observed in existing works, that led to the design of the proposed approach. Considering the topic of the paper, bio-inspired approaches like “Building a peer-to-peer information system in grids via self-organizing agents”, Journal of Grid Computing, 2008, should be considered.
- A model specifying the proposed methodology could be added to this paper.
- Add the organization of the paper and add Objectives of the paper in points before organization.
- Future scope of the approach should be highlighted and specified better
Author Response
Dear Reviewer,
Thank you for your insightful suggestions that help improve the understanding and scope of our work.
Their correct suggestions were taken into account in the corrections made.
Kind regards,
The authors.
Author Response File: Author Response.pdf
Reviewer 3 Report
The paper suggests a bio-inspired model for vowel recognition, which comprises the peripheral auditory system, the central auditory system, and the higher auditory system.
The last system was simulated with a simple neural network with a single hidden layer and a varying number of neurons in it. The authors carried out a series of computational experiments and concluded that the proposed bio-inspired model achieves higher levels of accuracy and sensitivity in vowel recognition. While the study shows promising results, there are a few major issues that could be addressed:
- The paper starts with the description of the hearing process and presents the motivation for building the bio-inspired model as the unclarity of the physiological mechanisms of functioning of the central and superior auditory system in terms of neural organization for the recognition of phenotypes. While the structure seems straightforward, it lacks related work which would locate the study around other open problems and alternative models and indicate a specific gap the authors intend to fill in.
- Related work may extend the list of references with other bio-inspired approaches to highlight the relevance and importance the current study, for example:
- Romera, M., Talatchian, P., Tsunegi, S. et al. Vowel recognition with four coupled spin-torque nano-oscillators. Nature 563, 230–234 (2018). https://doi.org/10.1038/s41586-018-0632-y
- Tan, H., Zhou, Y., Tao, Q. et al. Bioinspired multisensory neural network with crossmodal integration and recognition. Nat Commun 12, 1120 (2021). https://doi.org/10.1038/s41467-021-21404-z
- etc.
- The number of samples (5 males and 5 females) used to verify the model seems insufficient. At least it is difficult to evaluate the statistical significance of the provided results. Please consider providing the characteristics of the input vocalic signals. First of all, there is a need to present the number of observations and other properties which may affect the quality of neural network training and testing steps and result in the underfitting and overfitting issues).
- Before concluding the efficiency of the proposed model, it is important to assess the variances of accuracy, sensitivity, and specificity values given in Table 3. That would allow the authors to confirm the significance of the presented results with an appropriate statistical measure.
- Please increase the readability of Figure 2.
Author Response
Dear Reviewer,
Thank you for your insightful suggestions that help improve the understanding and scope of our work.
Their correct suggestions were taken into account in the corrections made.
Kind regards,
The authors.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The authors have not adequately addressed all the comments in the previous review including a performance comparison with other approaches in the scientific literature. I would also like to kindly ask the authors to provide point by answers to the issues raised in the review.
Author Response
Please, see the attachment
Author Response File: Author Response.pdf
Reviewer 2 Report
The bio-inspired concepts have to be introduced in the Introduction Section. A literature review is missing, then the authors have to introduce, also in a paragraph in the Introduction, some papers as the paper previously suggested. Another can be https://www.sciencedirect.com/science/article/abs/pii/S0167739X08000472
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 3 Report
I would like to ask the authors to provide a point-to-point response to each issue being raised in the review. Perhaps, the revised manuscript was mistakenly submitted as the author's response file but both files are needed to give the final recommendation.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Round 3
Reviewer 1 Report
I would like to thank the authors for the changes made.
Reviewer 3 Report
Good luck with further research
