Listen AI, an artificial intelligence solution that understands the meaning of sound

Detecting over 100 sounds and accurately distinguishes between noise and the target’s voice

Deeply, a technology startup, is researching solutions to analyze various sound sources and voices. It has developed ‘Listen AI’, a sound-based situation detection artificial intelligence solution that analyzes the shape of sound using deep learning and responds to various events and accidents. Listen AI detects various sound sources like the human ear and analyzes what the current sound means. It can detect various sounds such as screams, moans, commotion, vehicle accidents, explosions, gunshots, and more, and connects it to CCTV and dash cameras with edge devices. This creates a system that can detect situations and send information to the respective center for quick response. Listen AI can detect and customize over 100 types of sound, and can quickly integrate with and respond to devices in the control center via an application programming interface (API) and software development kit (SDK). Listen AI is unique in its ability to filter out even the most meaningless ambient noise.

Deeply CEO Lee Su-ji says “Unlike cameras, speech analytics has no blind spots, which is a big advantage, but it’s hard to filter out ambient noise. However, our solution can clearly distinguish between sounds coming from various media such as TVs and speakers, and emergency calls amidst the noise of everyday life. Because it can clearly distinguish between noise and target sounds, the false positive rate is very low and allows us to react immediately.” 

Boasting TTA authentication accuracy

Listen AI is TTA-certified, highly accurate and technologically advanced, with over 98% accuracy on F1 scores. Deeply has the world’s largest collection of nonverbal sound data, verified by more than 200,000 users worldwide, and includes 42 nonverbal sound datasets with more than 50,000 hours of audio. In addition to safety solutions, the innovative Listen AI provides machine solutions that detect abnormal sounds from machines or equipment to continuously monitor the entire process and equipment status. This innovative Listen AI is currently installed in 10 restrooms in the Incheon subway system, and in the Sejong Government Complex, it is also installed in places where CCTV cannot be installed, such as toilets. Deeply is also operating a system in collaboration with Korail that analyzes the sound of vehicles for malfunctions and abnormalities and predicts failures on high-speed rail, collecting all sounds generated when various parts malfunction and building them into big data. More recently, Deeply’s solution has also been applied to military drones, where it can be mounted on large unmanned drones and used by air traffic controllers to check for abnormalities. CEO Lee Su-ji published a paper at ICASSP 2024, an international conference on signal processing, in July of this year, and won the Director’s Award at the Digital Future Innovation Awards.