Introduction
For businesses developing medical technologies, R&D can begin with a practical technical problem.
A company may have a promising product concept or encounter limitations in product performance, manufacturing, software, or real-world application. However, the existence of a technical problem alone does not establish an R&D activity.
The more important questions are: What makes a technical problem suitable for R&D investigation? And how can an industry challenge be translated into a structured research problem?
Australia’s Research and Development Tax Incentive (R&DTI) supports eligible R&D activities undertaken by businesses seeking to resolve technical uncertainty through systematic investigation and experimentation.
Using voice restoration technology as an industry example, this article considers how businesses can move from a broad technical need to a more structured R&D framework involving research questions, hypotheses, experimental factors and validation methods.
Industry Background: Why Is Voice Restoration Difficult?
Helping people who have lost their natural voice regain effective communication has remained a long-standing challenge in medical technology.
Existing artificial voice solutions can provide basic communication support. Some devices, for example, use mechanical vibration or other assistive methods to enable speech generation. However, these approaches can involve limitations.
Generated speech may lack naturalness and personalisation and may not reflect the characteristics of the user’s original voice. Some technologies may also involve invasive procedures, adding complexity to long-term use and patient experience.
The industry challenge has therefore moved beyond: “How can a device help someone produce speech?”
A more complex question is: “How can technology help a person regain a voice that sounds more natural, more personalised and closer to their own vocal characteristics?”
Addressing this challenge may involve voice signal processing, artificial intelligence, software algorithms and medical device engineering.
Voice is also highly individualised. Where historical voice data or other user-specific information is limited, existing technologies may not provide a straightforward solution. As a result, businesses working in this area may need to investigate different technical approaches, test assumptions and progressively evaluate which methods can achieve the required outcome.
Industry Case: Laronix and Non-Invasive Bionic Voice Technology
Brisbane-based biotechnology company Laronix Pty Ltd provides a useful example of innovation in voice restoration technology.
Laronix has been profiled by the Australian Government in connection with the R&DTI and the development of its non-invasive bionic voice box technology.
Rather than focusing only on generating speech, public information about Laronix indicates that its technology development involves software-based approaches and artificial intelligence to support more natural and personalised artificial voices.
This type of technology presents challenges across several areas, including voice data processing, AI modelling, software systems and medical device applications.
The value of the Laronix example is therefore not simply that it represents an innovative product. It illustrates how a long-standing industry need can create a technical problem that requires continued research, experimentation and optimisation.
Importantly, the research framework below is illustrative. It demonstrates how a business working on a similar technical challenge might structure an R&D investigation and should not be interpreted as a description of Laronix’s specific R&D activities.
From an Industry Problem to a Research Problem
A technical challenge encountered during normal business operations does not automatically constitute an R&D project. The challenge first needs to be defined more precisely.
Businesses may need to identify:
What is technically unknown;
What outcome they are trying to understand or achieve;
What variables may influence that outcome; and
How different technical approaches can be tested and compared.
For voice restoration technology, a broad commercial objective such as “make the artificial voice sound more like the user” can therefore be developed into a more focused and testable research problem.
Potential Research Question
For a business exploring a similar technology, an illustrative research question could be: How can voice data analysis, artificial intelligence algorithms and software technologies be used to generate artificial speech that more closely reflects an individual’s vocal characteristics when available user-specific information is limited?
The research challenge is not simply to improve speech generation. It may require the business to investigate:
Which acoustic characteristics are most relevant to personalised voice generation;
How those characteristics can be identified and modelled;
How different technologies can reproduce or approximate those characteristics; and
How voice naturalness and personalisation can be balanced against practical system performance.
These questions provide a clearer basis for structured technical investigation.
Potential Research Hypothesis
An illustrative hypothesis might be: If key acoustic characteristics of an individual voice can be effectively extracted and modelled, and those characteristics can be incorporated into an AI-based voice generation process, then the resulting artificial speech may achieve greater similarity to the user’s original vocal characteristics than traditional mechanical voice generation methods.
The purpose of the hypothesis is not to predict that the proposed solution will succeed. Instead, it provides a technical proposition that can be tested through experimentation.
Different algorithms, data-processing methods and modelling approaches may then be compared to determine whether the proposed relationship is supported by the experimental results.
Potential Research Factors
A research team investigating this type of problem may need to evaluate several categories of technical factors:
Technical Input Factors: Quantity and quality of available voice samples; voice data processing and feature extraction methods; suitability of different voice analysis and generation algorithms; and software or model optimisation strategies.
Technical Performance Factors: Naturalness of generated speech; similarity to individual vocal characteristics; real-time processing capability; and overall system responsiveness.
Application Environment Factors: Physiological and acoustic differences between users; hardware system limitations; and reliability under real-world communication conditions.
Identifying these factors helps establish which variables need to be controlled, compared or evaluated during the research process.
Research Design & Methodology: Reducing Technical Uncertainty
For complex technologies involving artificial intelligence, voice processing and medical devices, the optimal technical solution may not be apparent from existing experience alone. A structured R&D process can help turn a broad industry problem into a technical question that can be systematically investigated.
For a voice restoration project, an illustrative research pathway may involve three stages:
1. Voice Data Analysis and Feature Modelling: Investigating which acoustic features have the greatest influence on personalised voice generation and how different data conditions affect the outcome. This stage helps identify the variables that should be tested during subsequent experimentation.
2. AI Model and Algorithm Optimisation: Comparing different algorithms, model structures and data-processing approaches to evaluate effects on voice naturalness, similarity to individual characteristics, processing performance, and system behaviour. The objective is to understand why different approaches produce different outcomes and revise hypotheses accordingly.
3. Application Environment Validation: Establishing performance measures to evaluate how the technology performs under representative operating conditions in practical use, identifying limitations, and informing further optimisation.
Through this process, a broad product requirement can be translated into a structured R&D pathway with defined research objectives, technical variables and validation methods.
Expected Outcomes & Industry Implications
In complex medical technology projects, the value of R&D is not limited to the creation of a final product. Experimental work can also generate new technical knowledge about which approaches are viable, which variables influence performance and why certain proposed solutions do not achieve the expected outcome.
For example, research may help identify which acoustic characteristics most strongly influence personalised voice output, which modelling approaches perform effectively under limited-data conditions, and where important technical trade-offs arise. This knowledge can strengthen internal technical capability, improve evaluation methods and provide a more systematic foundation for future product development.
The Laronix example illustrates how a real-world industry need can provide the starting point for technical investigation. However, businesses should distinguish between the broader product development project and the specific R&D activities undertaken within it. An innovative medical device, the use of artificial intelligence or a technically complex product does not automatically mean that every related activity qualifies as an eligible R&D activity under the R&DTI.
A well-structured R&D approach should clearly connect:
technical uncertainty → research question → hypothesis → experimentation → evaluation → new knowledge
This helps businesses explain not only what they are developing, but also what they are trying to discover, how the uncertainty is being investigated and what technical knowledge is being generated through the process.
This article provides general information only. R&DTI eligibility depends on the specific activities undertaken and the relevant circumstances. Businesses should assess individual activities against the applicable legislation and current Australian Government guidance.
Working on a project like this?
Tell us what you're building and the part you're not sure will work. A researcher from our Adelaide team (registered Research Service Provider RSP000047) reads every enquiry and replies within two business days. No documents needed.
