Bridging Speech and Hearing: How Audiology Advances Are Powering Next-Gen Voice Recognition
In many ways, voice recognition and hearing are closely connected. People with hearing loss often have to rely on devices to increase the volume and quality of the sounds around them to communicate and participate in daily life. The explosion in growth of AI-integrated voice recognition software for commercial use shows how machines are increasingly being asked to “listen” in complex environments. The technology is still falling behind human modes of hearing, but it is starting to catch up. These advances in audiology are poised to revolutionize the way that humans with hearing loss, as well as machines, are able to receive sound.
Current Challenges in Voice and Sound Recognition
Most people have had to manage issues with voice recognition technology, even if they do not have hearing loss. They speak to a chatbot, and the bot does not recognize what they said. Repeating phrases multiple times is common.
For people with hearing loss, similar changes occur with hearing devices. Most hearing devices, especially those relying on older technology, are not very effective at filtering out background noise or even understanding what the background noise is. When the technology cannot recognize what is being said, or which speaker to prioritize, the listener has more difficulty processing what they hear.
Benefits to Upgrades in Auditory Technology
Recent advances in auditory technology aim to improve the listening environment for machines as well as humans. For example, researchers at MIT used deep neural networks to make machines’ response to sound similar to how the human auditory system processes sound. These networks rely on a series of hidden layers between the input and the output, designed to improve the odds of accuracy.
To better mimic the human experience, researchers trained these networks on different varieties of auditory tasks, many of which included background noise. These advances create opportunities for audiology equipment suppliers and manufacturers to promote devices that receive sound more like humans do.
Modern Advances in Audiology
Here are examples of four emerging technologies that are shaping how machines and hearing devices interpret, filter, and respond to sound.
Mimic Inhibitory Neurons: For humans, the trick to filtering out background noise involves the use of inhibitory neurons. Neurons in the human brain are trained to listen for sound in different directions and at unique pitches. This allows the brain to focus its efforts in listening to the most important sounds. Inhibitory neurons block out some sound, acting similar to noise cancellation devices.
Research from Boston University shows that modern technology can mimic these neurons. The Biologically Oriented Sound Segregation System (BOSSA) uses a variety of cues, not just proximity and volume, to filter out less-important sounds and increase focus on a particular speaker. While still largely in the research phase, this technology could work to improve hearing devices and increase the accuracy of voice recognition.
Use Directional Audio to Filter: Voice recognition software must be able to recognize who is speaking in order to find the proper target for listening devices. Researchers have been working to filter out the wearers’ own voices to avoid transmitting hearing device users’ own statements back to them. The use of directional audio involves the placement of multiple microphones, one of which is aimed at the user. This microphone can help to recognize when the device user is speaking, to filter out that sound.
Voice recognition systems also benefit from directional audio by better identifying who is speaking in noisy environments, allowing systems to prioritize relevant input.
Predict Human Behavior: Humans act in ways that machines cannot, which is why the technology must tailor itself to human behaviors. The use of AI-integrated systems can provide more cues to help the technology meet the needs of users. Hearing devices are notoriously questionable at identifying the right speaker, often because they only have sound to process.
Some experimental hearing technologies can track the users’ eyes, which could help identify the most appropriate direction for the microphone—especially as people are unlikely to be listening to or engaging in conversation with someone standing behind them. Predicting these types of behaviors can help device manufacturers to customize technology to the way that humans really use it.
Follow a Conversational Rhythm: Although earlier examples of voice recognition technology focused on single, simple tasks, human hearing typically does not operate that way. People often speak to each other in conversations that follow a specific cadence. Earlier voice recognition systems focused on isolated commands, but modern research aims to capture conversational flow.
Researchers at the University of Washington found that they could train AI to recognize these patterns and provide accurate sound amplification at near-normal speeds. The technology relies on directional audio to identify the device user’s voice to filter it out and establish the rhythm of the conversation. Other speakers in the area who follow a similar cadence receive additional amplification and speed of transmission, while other noises are suppressed. The technology significantly improves speech clarity, posing potential improvements to hearing devices and other types of voice recognition and speech technology.
Building effective voice recognition software involves learning how humans speak and listen. Advances in audiology research are helping bridge that gap, improving sound processing for hearing devices and expanding the capabilities of voice-driven technologies.
Brian Frank is the marketing leader at e3 Diagnostics, a company that provides audiology equipment, service, and support solutions for hearing healthcare professionals. With more than six years at e3 Diagnostics, he focuses on leveraging data, marketing technology, and multichannel strategies to drive lead generation and revenue growth. Frank brings experience across Fortune 500 companies, small businesses, and the medical device industry, specializing in building high-performing campaigns and delivering measurable results.