Wearables business opportunities continue to grow. According to market research firm IDC, shipments of smart wearables will reach 240.1 million units in 2021, nearly double the market growth compared to 125.5 million units in 2017. The compound growth rate of the year is 18.2%; the positioning of wearable products is gradually becoming clear. At present, sports, positioning, calling, and payment are mostly used, especially in the fitness and healthy life, sports and fitness are more wearable. Important application areas of equipment, such as internationally renowned manufacturers Apple, Garmin, Samsung (Samsung) have launched clothing products focused on sports applications.
At the same time, the rapid rise of artificial intelligence (AI) also brings new applications to wearable devices, such as human motion recognition, through deep information collection and analysis, and AI wearable devices will provide more effective and accurate data. Let the consumer refer to it; and the sensor plays a key role in this wave of exercise fitness and AI wave. The quality of the message collection directly affects the subsequent processing and application. Therefore, all major sensor suppliers are currently working to improve the accuracy of the sensor. degree.
AI/Multi-function application promotes sensor accuracy
The rapid development of AI has driven various innovative applications to spring up, and this wave has spread to wearable applications. Chen Jiancheng (Figure 1), product marketing manager of STMicroelectronics Asia Pacific, said that with fitness, medical care Diffusion, coupled with the rise of AI, the function and application of wearable devices is becoming more and more extensive; and in order to effectively analyze the increasingly large data and provide more accurate and trustworthy analysis results for consumers, the back-end algorithms become more complicated. In order to reduce the algorithm operation time, equipment manufacturers also begin to demand sensor accuracy. In other words, following small size and low power consumption, accuracy has become one of the important considerations for sensor design.
Li Dinghan, general manager of Ams Taiwan, also pointed out that wearable devices will be developed for specialized products, such as professional sports watches, medical care equipment, etc., and in the face of more extensive applications and more complex algorithms, sensors such as 2000 Accuracy can be improved. The more accurate the data is detected in the first time, the more the latter algorithm can analyze the correct data for the consumer's reference. This will help the terminal manufacturer to speed up the product and application development time. This is the main factor that the accuracy of the sensor is constantly improving.
Improve sensing accuracy Three parameters are key
The AI trend has swept the world, and the application of wearable devices has been increasing. The complexity of the algorithms behind it has increased, which has also led to the improvement of sensor accuracy. Chen Jiancheng pointed out that to improve the accuracy of the sensor, three key parameters should be referenced in the design. , respectively, noise, stability, and tolerance (Tolerance).
The above three are the key parameters to enhance the accuracy of the sensor, and each of these parameters contains many adjustment items, such as vibration rejection, Flicker Noise, and high noise. Frequency (High Frequency); Stability is the need to pay attention to the stability of the (Stability of Time), Stability vs Temperature (Resistance vs Temperature), Reliability (Repeatability); As for the error is the offset (Offset), sensitivity (Sensitivity) and nonlinear (Non-Linearity).
Chen Jiancheng further explained that noise, stability and error parameters are important parameters for adjusting the accuracy of the sensor. If the noise is not adjusted, it will easily lead to distortion of the detected data. Stability is to ensure that the product can respond to different environmental changes and continuously collect accurate data. .
For example, Chen Jiancheng, if a sensor has a vibration of 10 GHz, it means that the sensor collects information under 10 GHz vibration; but if the sensor is not stable enough, its vibration frequency does not change with the user's movement or external environment. Jumping from 10GHz to 20GHz and then jumping back to 10GHz from 20GHz means that the data collection of this sensor is non-linear, so the data collected is quite inaccurate, and it is easy to cause the algorithm to run smoothly, making wearable device related applications. Analyze distortion, or stop. Therefore, the stability of the sensor is highly important for accurate detection.
As for the error value part, there is no doubt that the smaller the error value, the better. At present, the supplier's target of each sensor is to pursue the error value within plus or minus one. Chen Jiancheng said that the current industry players are actively pursuing the error value. As mentioned above, the algorithm of the back end is more and more demanding on the accuracy of the sensor. It is hoped that the data collected by the sensor in the first step of detection will be quite accurate, so as to reduce the operation and adjustment time after the algorithm.
Chen Jiancheng explained that the reason why the wearable device manufacturers did not specifically require sensor accuracy is because the function of the wearable device is relatively simple, and the demand for data analysis is not so high; but with the rise of AI, the era of big data is coming, plus Consumers are increasingly demanding the functionality of wearable devices. Manufacturers must develop more innovative applications through algorithms, making the algorithm more complex. In this case, if the sensor is not accurate enough, it can't be the same as in the past. The back-end algorithm takes a little more time to adjust for the lack of hardware. In summary, small size and low power consumption are now a must for sensors, and wearable device manufacturers are beginning to have sensor accuracy. As required, I hope to accelerate the product and application development time.
Health management prevails Optical sensing components re-evolution
As consumer demand for wearable devices increases, Wang Weidi, technical associate of the Sony Design Center of Roma Semiconductors, believes that this trend will drive the development of wearable devices towards a two-polar (lower-order, higher-order) market, while high-order The more features the product has, the higher the accuracy requirements for the sensor.
Such as detecting running, swimming, how many calories burned, walking frequency, etc. These applications rely on sensors to collect data, and then analyzed by the back-end algorithm. If the accuracy of the sensor is improved, it will help the algorithm to analyze the data more. accurate.
Wang Weidi takes the application of blood pressure detection as an example. It points out that consumers' requirements for related applications are not only limited to blood pressure, pulse beat times, etc., but also hope to learn physiological signs such as blood vessel age. To achieve this goal, application Sensors for blood pressure detection must increase the sampling frequency and increase the number of measurements per unit time in order to obtain complex physiological signs.
As a result, ROHM continues to improve the process technology of its sensors, such as the company's recently introduced optical pulse sensing IC, which supports pressure measurement and blood vessel age measurement, the BH1792GLC, whose photoreceptor consists of an infrared cut filter and a Green filter. The composition, through the Green filter, allows only the green wavelength to pass through, thereby reducing the interference such as infrared rays and infrared light, so that the sensor is in a strong environment such as strong shaking during movement and strong infrared rays such as sunlight. Can perform high precision pulse measurement.
In addition, the product can also be used for high-speed sampling at 1024 Hz. Compared with traditional products, the pulse measurement speed is up to 32 times faster, which can support the current measurement of the vital signs such as pressure, blood vessel age and other vital signs.
Wang Weidi pointed out that, of course, in addition to the above-mentioned methods used by the new pulse-sensing ICs, there are many aspects to consider the accuracy of sensor enhancement, such as how to optimize the process, make the internal capacitance of the sensor smaller, or reduce the board. The inductance between the upper traces, etc., makes the sensor more accurate, which is a place to pay attention to in the hardware design.
On the other hand, the future health care will become an important application market for wearable devices. Multi-party cooperation and formation of an industrial cooperation chain must be the only way for future sensor suppliers. Wang Weidi pointed out that when the sensor detects the data, the back end Data acquisition, analysis is related to the algorithm. To this end, the sensor supplier must cooperate with the algorithm company in addition to the hardware, soft and hard to improve the overall detection performance of the sensor; as for data interpretation, it depends on the professional The medical team supports, in this way, through the cooperation of many parties, can provide consumers with trustworthy and accurate data and analysis results.
For the future development of wearable devices, Li Dinghan also holds the same view that the market segmentation of low/high-end will become more and more obvious, and high-end wearable devices will be developed for specific applications, such as medical care, sports, etc.; It is not like the past, just a simple heartbeat, walking step detection, but to interpret the user's current state of motion, or analysis of exercise results.
Therefore, the role of the sensor in the wearable device is becoming more and more important. However, the use of the wearable device has always been unfixed. Each user's skin color and hair volume are different, and there are many variables. Therefore, how to improve the sensor accuracy. It has become a problem for all suppliers now. As the accuracy of the products developed is higher, the more accurate the detected data, the faster the terminal manufacturer's product and application development time can be accelerated to meet consumer demand.
So, how do you further improve the accuracy of the sensor? Li Dinghan takes the sensor that measures blood pressure as an example. It is pointed out that most of the current use of optical sensors to detect blood pressure, but as mentioned above, each person's physiological characteristics are different, so when measuring There are always many variables. Therefore, the industry must combine light sensors with other technologies to enhance the separation performance of the light sensor for invisible light, so that it can eliminate many external factors (such as noise, light source) in the first time. Path changes), allowing the back-end algorithm to analyze data data faster.
For example, ams recently introduced the new new vital sign sensor, the AS7024, which integrates all the hardware components including the AS7024 and performs blood pressure measurement, heart rate measurement (HRM), heart rate variability (HRV) and electrocardiogram (ECG). The sensor IC consists of three LED lights, a photodiode, an optical front end for HRM and program equipment, and an analog front end for ECG, all integrated in a 6.1mm × 2.7mm micro package. Through the external sensor interface, the reference design can also be used for the calculation of vagal tone and arterial elasticity, as well as skin temperature and skin resistivity.
In addition, the AS7024 HRM operation is based on photoplethysmography (PPG), which measures the pulse rate by the sampled light adjusted by the blood vessel. When the blood pulse passes, the blood vessel expands and 7543 shrinks. This technique is Verified. Electrocardiogram is the standard method for measuring the electrical impulses generated by the cardiac sinus node.
Li Dinghan said that in the process of designing sensors, it must also work closely with the module manufacturer or equipment manufacturer. Because each manufacturer designs products with different functions, their customers and users are different. There are many kinds, some are water repellent, some require three meters of water underwater, and some hope to be 10 meters underwater, each product design idea is different; and any factor in design, such as coating, The placement of the sensor, the angle of incidence of the light, etc., will affect the accuracy of the sensor. Therefore, the sensor supplier must also cooperate with the terminal manufacturer to improve the accuracy of the sensor according to its hardware design environment and ideas.
Compact/precise/low-power sensor-required features
NXP Semiconductors Greater China Marketing Manager Mr. Fang pointed out that in 2018, the demand for many wearable applications has grown significantly, such as smart watches that support mobile payments, wristbands; easy to use small headphones and VR glasses Etc.; Portable health monitoring and medical care products.
Among them, in portable health care equipment applications, consumers require measurement accuracy to achieve medical-grade performance levels, family and doctors can instantly obtain and monitor the most reliable health data; or wearable fitness equipment For example, most devices on the market now have simple step counting capabilities, and the more complex sensor and sensor fusion algorithms can reduce the error to 2%, making wearable fitness equipment not only for simple step counting, but also Identify different motion states, as well as determine posture anomalies.
Therefore, in the future, in addition to the small size and low power consumption of the sensor, the sensor must be continuously refined, so that the sensor is not easily affected in a more severe environment or a wider operating temperature range; Machine learning and AI algorithms for richer, smarter application scenarios and services.
Wang Weidi added that for consumers, wearable devices are not 'necessary' or 'need' products in life, but they will be 'want' products; therefore, how to add more functions to wearable products to attract consumption It is a constant strategy for wearable device manufacturers to purchase products. Therefore, functions such as detecting heartbeat, number of steps, electrocardiogram, blood oxygen, respiratory rate, etc. will be more and more, and these applications are all behind Algorithm support. Especially after the rise of AI, these applications are more and more. It is conceivable that in the future, when the wearable device is combined with AIOT, the business opportunity will explode. In response to this trend, the accuracy of the sensor must be Continue to improve, in response to more complex algorithms.
All in all, the rapid rise of AI brings new applications to wearable devices, and in response to more complex algorithms, the sensors responsible for collecting information are becoming more and more important, not only with small size, low power consumption, but also with precision. With the upward climb, it will not cause poor quality of message collection, but will affect the subsequent algorithm processing and application, and further bring consumers a better experience.