January 13, 2021
Telink’s high-performance, low-power consumption chips empower manufacturers to deliver always-on local voice recognition capabilities across a variety of devices.
As we settle into 2021, we may not have flying cars or readily available virtual reality systems, but we do have access to one staple of 20th-century visions of the 21st century: voice-controlled technology. Over the last couple decades, talking to one’s technology — and having it respond — has moved from the realm of science fiction to an everyday reality. In fact, over half of American adults have used voice commands at some point in their life, and of these adults, roughly a quarter do so on a daily basis.
In many ways, convenience is the heart of voice technology’s appeal. The ability to speak commands to a smartphone, smart speaker, or personal computer or smart television (via a voice-enabled remote) instead of having to type out commands saves users a great deal of time and facilitates multitasking.
Initially, technological limitations forced voice-enabled technologies to rely on push-button voice and cloud processing. To activate their devices’ voice capabilities, users would have to push (and often hold) a button before speaking their commands. Their voice command would then be transmitted to the cloud, where it would be run through voice recognition software. Once the software determined the proper command, it would send the command back to the voice-controlled device via a local gateway. While this cloud-based approach works well enough, it can result in slow processing speeds (since voice commands have to travel to the cloud and back) and a subpar user experience. Generally speaking, it also limits users to making one voice command at a time.
However, recent advances have enabled a new approach to voice that delivers greater convenience: always-on local voice. Instead of waiting to be activated by the push of a button, a device equipped with always-on local voice capabilities remains on in a passive state at all times awaiting a designated “wake phrase” or “trigger word.” With this approach, even if a device is on standby, the device’s microphone will listen for its wake phrase and spring into action when a user utters the phrase — think, for example, “Hey, Alexa.” Moving this part of the voice recognition process to smart devices themselves reduces processing time and enables users to execute multiple voice commands simultaneously. This makes always-on local voice an ideal choice for a wide variety of devices, but especially for simple smart devices and devices that do not need to be connected to the internet.
In order to make always-on local voice a viable approach for wireless devices, device manufacturers need chips like Telink’s that can help them overcome two key challenges: power consumption and environmental noise. As its name suggests, always-on local voice requires a device’s microphone to be listening for the device’s wake phrase at all times. While this does not involve highly power-intensive tasks — the microphone simply captures short stretches of audio that are then processed locally (that is, not transmitted to the cloud and back) to identify the wake phrase — it does drain an ordinary device’s battery rather quickly.
For a device like a smart speaker that can easily be plugged in to an AC power source at all times, this is not a significant issue, but for always-on local voice use cases like smartphones, hearables, and wearables, it has been a major challenge until the last few years. Fortunately, Telink’s ultra-low power consumption chips empower manufacturers to build devices using standards like Bluetooth Low Energy that are capable of delivering always-on local voice capabilities without compromising battery life.
Additionally, devices that are built for portability present much more complex, much noisier sonic landscapes. Whereas a smart speaker will usually remain in a relatively quiet setting like a home or a closed-door office, smartphones, hearables, and wearables need to be able to deliver their always-on local voice capabilities in noisy environments like coffee shops, train platforms, and gyms. As such, these devices must be able to pick up their wake phrase amid a continuous stream of ambient noise.
In order to achieve such “selective listening,” a device needs to be equipped with a range of speech preprocessing capabilities. Telink’s RISC-V-powered TLSR9 Series offers advanced noise suppression, hybrid acoustic noise cancelation, and acoustic echo cancelation capabilities that ensure devices’ microphones only hear what they need to hear — namely, their wake phrase.
As chips like Telink’s have made it increasingly viable for manufacturers of all types, always-on local voice has started to appear in a growing variety of devices. Some of the most common use cases include remote controls, hearables/wearables, smart home devices, and in-vehicle infotainment systems.
Telink’s low-BOM cost, energy-efficient, voice-enabled chips help device manufacturers deliver the always-on local voice capabilities consumers demand. Our chips pack everything manufacturers need into a compact, high-performing package that can be customized to support all popular communications standards.
We have developed a range of multimode solutions for always-on local voice that deliver exceptional performance and unparalleled flexibility at an unbeatable price point. To learn more, contact us today.