Shulou(Shulou.com)11/24 Report--

CTOnews.com, March 19, according to the official website of the College of Integrated Circuits of Tsinghua University, Professor Ren Tianling and his team of the School of Integrated Circuits of Tsinghua University have recently made important progress in intelligent voice interaction. The wearable artificial larynx developed can sense multimodal mechanical signals related to laryngeal sound for speech recognition, and rely on thermoacoustic effect to play the corresponding sound. The research results provide a new technical approach for speech recognition and interaction system.

▲ uses intelligent wearable artificial throat for silent voice interaction | Picture Source: voice on the official website of the College of Integrated Circuits of Tsinghua University is an important way for human communication, but the speaker's health status (such as sound disorders caused by neurological diseases, cancer, trauma, etc.) and surrounding environment (noise interference, transmission media) often affect sound transmission and recognition. Researchers have been improving speech recognition and interaction technologies to cope with weak sound sources or noisy environments. Multi-channel acoustic sensor can significantly improve the accuracy of sound recognition, but it will lead to larger equipment volume. On the other hand, wearable devices can obtain high-quality raw voice or other physiological signals. However, at present, there is no sufficient evidence to show that the movement pattern of laryngeal muscle and the vibration of vocal organs reflected in the body surface imply recognizable speech features, and there are no experiments to prove its completeness as a speech recognition technology.

To solve this problem, Ren Tianling team developed a graphene-based intelligent wearable artificial larynx (AT). Compared with commercial microphones and piezoelectric films, artificial larynx is highly sensitive to low-frequency muscle movement, intermediate-frequency esophageal vibration and high-frequency sound waves, as well as anti-noise speech sensing. The perception of the mixed mode of acoustic signal and mechanical motion enables the artificial throat to obtain a lower speech fundamental frequency signal. In addition, the device can also realize the sound playback function through the thermoacoustic effect. The production process of artificial larynx is simple, stable and easy to integrate, which provides a new hardware platform for speech recognition and interaction.

The team also uses the artificial intelligence model to recognize and synthesize the signals perceived by the artificial larynx, realizing the high-precision recognition of basic speech elements (phonemes, tones and words), as well as the recognition and reproduction of fuzzy speech in patients with laryngeal cancer. it provides an innovative solution for the communication and interaction of people with voice disorders.

The experimental results show that the mixed-mode speech signal collected by the artificial throat can recognize the basic speech elements (phonemes, tones and words), and the average accuracy is 99.05%. At the same time, the anti-noise performance of the artificial throat is obviously better than that of the microphone, and it can still maintain the recognition ability under the ambient noise above 60dB. Ren Tianling's research team further demonstrated its voice interactive application. By integrating the AI model, the artificial larynx can recognize the daily words vaguely spoken by a laryngectomy patient with an accuracy of more than 90%. The recognized content is synthesized into voice and played on the artificial throat, which can initially restore the patient's speech communication ability.

According to reports, the artificial throat still has a lot of room for optimization and expansion, such as improving sound quality and volume, increasing voice diversity and expression, and combining other physiological signals and environmental information to achieve more natural and intelligent voice interaction. The research team hopes to benefit more people with voice disorders and voice interaction users through further research and cooperation.

CTOnews.com learned from the website of the College of Integrated Circuits of Tsinghua University that the results, entitled "mixed-mode speech recognition and interaction using wearable artificial throats", were published online in Natural Machine Intelligence, a sub-journal of Natural artificial Intelligence, on February 24.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.