Preferred learners
- Anywhere
- Academic experience
Categories
Project scope
- What is the main goal for this project?
-
We make a smartphone app called HeardThat which uses AI to enhance speech so that the app user can hear more easily in noisy areas. Understanding speech in noise is the most common complaint of adults with hearing loss, even if it is mild loss, and whether or not they have hearing aids. Hearing loss affects about 20% of the world's population.
Our HeardThat product has been used over a million times. It uses the microphones in a person's smartphone to gather environmental sound and then uses AI running on the phone to remove the noise.
To increase the number of use cases where our technology can be used, we have recently introduced the HeardThat Remote Mic Kit (RMK). It consists of two wireless clip-on microphones and a receiver that plugs into the phone.
The AI algorithms that HeardThat uses have been optimized for audio coming from a phone’s built-in mics. The mics in the RMK have somewhat different characteristics. The goal of this project is to gather data from RMK mics and use it to fine tune the existing AI models.
A stretch goal is to investigate synthesizing data for training the models. Such data needs to be as realistic as possible to match the characteristics of the mics and room effects (such as reverb). The goal is to recommend the best approaches.
- What tasks will learners need to complete to achieve the project goal?
-
Data collection
- This amounts to taking a few phones running our app to noisy locations like coffee shops, putting the app into record mode, and chilling out while it records the ambient background noise.
- A secondary activity is to record people talking while using the app with the RMK. This will include listening in real time and observing the effectiveness of the noise removal.
Assess and improve model quality
- Using the collected data, establish a baseline for model quality using our existing models.
- Fine tune the existing models using the collected data to improve the performance when using RMK mics.
Assess data synthesis techniques (if time permits)
- Assess different ways of doing the synthesis to select the best approaches.
Deliverables
- Collected data
- Any related project artifacts such as code, GitHub issues, etc.
- A report describing methodology and conclusions
- What is the main goal for this project?
-
We make a smartphone app called HeardThat which uses AI to enhance speech so that the app user can hear more easily in noisy areas. Understanding speech in noise is the most common complaint of adults with hearing loss, even if it is mild loss, and whether or not they have hearing aids. Hearing loss affects about 20% of the world's population.
Our HeardThat product has been used over a million times. It uses the microphones in a person's smartphone to gather environmental sound and then uses AI running on the phone to remove the noise.
To increase the number of use cases where our technology can be used, we have recently introduced the HeardThat Remote Mic Kit (RMK). It consists of two wireless clip-on microphones and a receiver that plugs into the phone.
The AI algorithms that HeardThat uses have been optimized for audio coming from a phone’s built-in mics. The mics in the RMK have somewhat different characteristics. The goal of this project is to gather data from RMK mics and use it to fine tune the existing AI models.
A stretch goal is to investigate synthesizing data for training the models. Such data needs to be as realistic as possible to match the characteristics of the mics and room effects (such as reverb). The goal is to recommend the best approaches.
- What tasks will learners need to complete to achieve the project goal?
-
Data collection
- This amounts to taking a few phones running our app to noisy locations like coffee shops, putting the app into record mode, and chilling out while it records the ambient background noise.
- A secondary activity is to record people talking while using the app with the RMK. This will include listening in real time and observing the effectiveness of the noise removal.
Assess and improve model quality
- Using the collected data, establish a baseline for model quality using our existing models.
- Fine tune the existing models using the collected data to improve the performance when using RMK mics.
Assess data synthesis techniques (if time permits)
- Assess different ways of doing the synthesis to select the best approaches.
Deliverables
- Collected data
- Any related project artifacts such as code, GitHub issues, etc.
- A report describing methodology and conclusions
- How will you support learners in completing the project?
-
Students will connect directly with us for mentorship throughout the project. We will be able to provide answers to questions such as:
- Our current products and applications of AI / ML
- The current data set and guidance in assessing
- Current industry standard approaches to AI / ML in our application area
- Input on choices, problems, or anything else the students might encounter.
- Guidance on the software tools and techniques to be used
- Access to our product software and hardware
- What skills or technologies will help learners to complete the project?
-
In order to complete this project, students can self-teach, but it is beneficial to be familiar with:
- Modern ML techniques and data analysis
- Software development using ML languages like Python
- ML ecosystem tools such as Jupyter Notebook
Again, students can complete this project without any of the prior knowledge listed here. Students will be expected to research and learn more about the above as the project goes along.
Supported causes
About the company
- https://heardthat.ai
- 2 - 10 employees
- Technology, Hospital, health, wellness & medical
Singular Hearing develops new approaches for hearing assistance based on AI and machine learning. We have extensive experience in speech and audio technology. Our first product is HeardThat, an award-winning smartphone app that turns smartphones into sophisticated hearing-assistive devices that can help anyone who has difficulty hearing speech in noisy environments.