The demand for voice recognition is strong, ST and Infineon are rushing to launch their own solutions

The voice recognition market is booming. According to market research firm Strategy Analytics, by 2022, it is estimated that there will be more than 62 million devices with personal voice assistants in the global consumer market. To gain a foothold in this market, the semiconductor industry has launched a new generation of solutions. For example, Infineon combines radar, MEMS microphones and audio processors to further improve the voice recognition performance of MEMS microphones; STMicroelectronics (ST) has joined hands with Sensory, a developer of voice interface and keyword detection algorithms, and DSP Group, a supplier of wireless chipset solutions, to jointly develop efficient voice detection and processing microphone technology.

Lu Peiru, an analyst at the Electronics and Systems Research Group of IEK, ITRI, said that although voice assistant-related services are still in their infancy, they will penetrate into smart home appliances, car systems, and even more IoT devices in the future as market demand grows rapidly. Currently, semiconductor manufacturers, telecom operators and other companies that want to enter the voice recognition market are not only actively developing related applications, but also focusing on how to improve voice recognition performance.

To this end, semiconductor companies are accelerating their layout. Infineon combines radar, silicon microphone sensors, and XMOS audio processors to provide far-field voice capture capabilities through audio beamforming and radar target presence detection technology, ensuring that a wide range of voice control devices can perform ideal sound recognition to perform digital voice assistance functions, thereby improving the dilemma of MESM microphones being unable to accurately identify the location of the voice source when multiple people are talking, and unable to separate the noise from the object.

It is reported that Infineon's 60GHz 2Tx/4Rx radar IC, coupled with the accompanying antenna and 70dB SNR microphone, will help overcome these obstacles. This microphone uses the company's dual-backplate MEMS technology for far-field voice capture and beamforming. In addition, the microphone's improved SNR will further enhance performance; and the XMOS audio processor will analyze the signal data from the company's digital microphone array and adjust the angle and distance data provided by each microphone to form a beam at the angle identified by the radar data. Through the combination of radar and XMOS beamformer, the microphone can accurately lock onto a specific object even if the object is moving and there is ambiguous noise.

On the other hand, STMicroelectronics has teamed up with DSP Group and Sensory to release the technology of efficient voice detection and processing microphone. This component integrates ST's low-power MEMS microphone, DSP Group's ultra-low-power voice processing chip, and Sensory's voice recognition firmware in a micro system-level package (SiP), and achieves lightweight packaging, longer battery life, and advanced functions through the company's advanced packaging technology.

At the same time, the new microphone uses DSP Group's HDClear low-power audio processing chip, which can significantly reduce energy consumption and extend the battery life of battery-powered devices for many years without the need for charging or replacing batteries; and the system executes commands immediately without the need to pre-identify commands, and the response speed of voice commands will be faster.