Using AI to improve speech quality
The AI-powered V-Series Gen II control units provides unparalleled audio in extreme environments.
As a market leader in tactical communication systems, INVISIO is relentlessly pursuing
innovation to ensure high-quality solutions. Its new AI-powered technology platform and control units, V-Series Gen II, provides unparalleled performance in extremely noisy environments. The latest generation system builds on INVISIO’s strong legacy of products and technology, proven by thousands of operations worldwide, and constantly refined thanks to feedback and research around customers’ needs and requirements.
A key area that has allowed INVISIO to make a leap forward in improving sound quality is through the use of AI. Artificial intelligence has transformed life in the 21st century, permeating through every sector and offering the opportunity to redefine problem-solving and find otherwise elusive solutions to certain situations.
INVISIO has used AI to achieve a 30% increase in speech quality via its new generation of tactical communications systems, and at the same time managed to run the AI-algorithms at low power.
AI has allowed INVISIO to push boundaries and reduce sound degradation and unwanted noise to a minimum. This ensures crystal clear communication in mission-critical environments and reduces the cognitive burden for the user to allow a better, safer experience.
Figure 1: A mix of bone conducted speech and surrounding noise is picked up by the control unit and transformed into a high-quality speech signal that is transmitted over the radio network without degradation of speech intelligibility.
The adoption of AI
The use of artificial intelligence allows the analysis of large amounts of data to first identify the obstacle and then solve it, without the limits of human capability to create a defined task or specific algorithm.
INVISIO recruited experts, who were working on reconstructing bone-conducted speech at Danish Technical University (DTU), the same technology used by INVISIOs X5 in-ear headsets. They had ascertained the need for AI to provide an enhanced solution. “They collected a range of data from recordings of people speaking into INVISIO’s X5 headset with its bone conduction microphone as well as high quality microphones, producing both degraded, lower quality speech and high-quality speech,” Jan Larsen, Sr. VP of R&D at INVISIO, told Soldier Modernisation. “By inputting this data, the team used AI to create its own solution to reduce any degradation to a minimum and provide even higher quality transmissions.”
The next step was to collect more data in order to deploy the algorithm in mission-critical environments. To do so INVISIO’s experts recorded speech from multiple nationalities and different genders as well as including data involving a range of noise environments, mixing clean speech segments with a database of noises from machine noise to gunshots and helicopters. The use of this larger data set allowed the development of a stable product that could be deployed into real-world situations.
But it didn’t stop there. “AI may have improved sound quality, but also brought its own challenges,” said Larsen. “It demanded more power compared to our low-power headsets and control units, which meant the technology had to be reduced in size so it could be used practically in the field.” INVISIO managed to reduce the power required by a factor of 80 times. The ability to embed the AI itself in the offline control unit also provided added reassurance in terms of security, protecting communications from cyberattacks and any compromise.
Figure 2: Up to 30% improvement was achieved when measuring speech quality in relation to noise and comparing the new AI-powered V-series Gen II and the X5 headset with the existing V-Series.
The results speak for themselves
INVISIO’s analysis of the results of the AI using the Perceptual Evaluation of Speech Quality (PESQ) - an accepted family of standards used to assess speech quality - revealed clear improvements to speech quality. They showed a 30% increase in quality - crucially improving users’ ability to perceive speech in critical situations.
In addition, a large field test involving 400 users produced anecdotal evidence suggesting that the use of the AI within the systems helped reduce cognitive burden for users. INVISIO collected feedback from the users via questionnaires in which they were asked about their experience. Comments within the feedback showed that users heard transmissions more clearly than with previous in-ear headsets, while the AI also meant that those broadcasting transmissions did not have to repeat themselves.
Developments in the future
The use of AI in the new generation of INVISIO technology and its ability to increase speech intelligibility and remove unwanted noise has allowed a step-change in developments to tactical communication. For Larsen, the potential is clear, with research already providing the stimulus for further improvements to the algorithm, as well as the potential development of new firmware for control units. Possibilities also include the potential to expand the use of AI to other potential uses within INVISIO’s area of expertise.
“The introduction of AI has enabled us to make a huge leap forward. We’ve achieved a 30% increase in speech quality as well as finding a solution for power consumption, allowing us to reduce the AI algorithm by a factor of 80. This paves the way for future developments and opens the door to even more improvements and solutions in mission-critical communication. We look forward to exploring these so we can continue to develop market-leading communication technology for the defence sector.”
For more information please visit: