CEVA Technology Symposium 2016 was held in Hsinchu, Taiwan on November 21, 2016 and two more sessions were held later in Shenzhen and Shanghai respectively. Each year CEVA holds a series of annual events in Asia with growing numbers of participants and growing reputation. This is the 10th year of the symposium series and it has become an important annual technology event in the semiconductor industry that are highly regarded by semiconductor firms and IC design houses.
CEVA is the leading firm for silicon intellectual property (SIP) in digital signal processing (DSP). The firm's core technology provides memory bandwidth for data processing and unique DSP engines. CEVA chips have been installed in numerous types of smart devices, and today more than 90% of the mobile phones have some CEVA technology inside. Through many years of hard work in developing the DSP core and building a comprehensive semiconductor ecosystem, CEVA has successfully assisted many IC design houses around the world to speed up product development, shorten time-to-market and expand their markets.
CEVA Technology Symposium's theme this year was about the development trend and direction of DSP applications. It was a technological feast with discussions on Internet of Things (IoT) smart devices and machine learning applications that use connectivity and neurological technologies. Without a doubt, the rise of new types of smart devices and computer vision has promised to make IoT applications to a main growth engine for the semiconductor industry. Technologies such as computer vision, face recognition, ultra-high resolution, scene detection, machine learning, artificial intelligence (AI), image formation and visualization allow imaging devices to feature innovative designs.
CEVA and its partners displayed products and applications developed using the CEVA platform at the interactive booths at the Hsinchu symposium, which also saw participation of semiconductor firms and IC design houses from China. China's semiconductor industry has been growing rapidly. With increasing investments and major mergers, the development in China is providing strong stimulations to the global semiconductor industry. The Hsinchu symposium featured a keynote speaker from China-based Brite Semiconductor – a firm invested by China-based Semiconductor Manufacturing International Corporation (SMIC) – to introduce China's semiconductor manufacturing and design processes, giving the participants a glimpse of some other options for system on a chip (SoC) solutions with which they could speed up the development of IoT applications.
IoT and computer vision to shape the future mainstream applications of DSP technology
CEVA Technology Symposium 2016 arranged keynote speeches in the morning and two technology sessions in the afternoon that respectively discussed the topics of IoT and computer vision, exploring issues concerning electronic devices' connectivity functions and smart capabilities as featured in imaging, vision, audio, sensor, connectivity and AI applications. CEVA invited IC design houses such as NextG-Com, Sensory, ArcSoft, Corephotonics, Almalence, emza Visual Sense, StradVision, Socionext, and Pilot.ai, and their partnering firms to jointly display their products and applications that use CEVA's DSP technology.
Smart IoT devices head toward wide-area communication technology
The increasing popularity of IoT and Big Data analytics-enabled smart devices means smart IoT devices have to have strong connectivity abilities that support different industry standards such as 3G/4G baseband, WiFi, Bluetooth BLE and Zigbee. Different products supporting various connectivity technologies have been introduced for different market segments. In addition to frequently-renewed radio frequency (RF) solutions for WiFi and Bluetooth, LTE products that are low in power loss and costs with longer range of connectivity have been the focus of CEVA and chipmakers.
Electronic sensors and smart IoT devices that are connected to one another are estimated to reach 50 billion units around the world in 2020. Therefore, competition is expected to heat up among different transmission standards. It is foreseeable that there will be a lot of smart products for indoor or smart home solutions competing for market share. One single device will need to support multiple network connection technologies such as a combination of WiFi, Bluetooth and Zigbee, and a key to the device's success will be lower cost and power loss.
Short distance connectivity is important but as more IoT devices are heading outdoors, long-distance connectivity applications have been emerging with the introduction of many devices for smart farming and long-distance transportation. As demand for long-distance connectivity has been growing, 3GPP stepped up efforts standardizing the three major low-power loss long-distance wireless network standards: NB-IoT, LTE-M, and EC-GSM. The standards were completed in June 2016.
Smart IoT devices with lower-power wide-area telecommunication technologies are expected to show an impressive cumulative annual growth rate (CAGR) of 30% to reach around 2.1 billion units in the market by 2022. The applications will include smart cities, smartwatches, smart hydrometers and gas meters, and smart vending machines. This is seen as an important engine driving strong growth in the next few years. CEVA expects wide-area smart IoT devices to become a success.
Computer vision increases machine learning efficiency pushing image recognition applications to grow rapidly
To improve the "smartness" of electronic devices, the crown jewel of the DSP technology application is computer vision and image recognition. DSP core used to process the input and analysis of large amounts of image data for automobile solutions in pedestrian detection and parking space tracking has attracted much media attention.
In the consumer electronics market, many leading smartphone makers adopt dual-lens cameras – one lens with 1x standard wide angle and the other 2x optical zoom. The dual-lens camera will require more capabilities of image processing and computing algorithm to support growing and wider adoption of visual applications. Hence DSP technology has become irreplaceable. Whether it is smart home, security monitoring, self-driving vehicles or drones, the smart functions require high-density algorithms to analyze the specific features of different images. DSP technology plays a leading role in image and voice recognition.
From the point of view of image resolution, HD, UHD, 4K and even 8K streaming of videos with frame rates from 30fps to 60 fps has become mainstream. The challenge is that large amounts of data have to be processed and stored. The DSP core from the fifth-generation CEVA XM6 has enhanced processing abilities and achieved low power loss, which is important to mobile devices. As a result of joint efforts with partners through the design and application of 3D images and neural network systems, the product has higher smart recognition quality for future smart applications.
CEVA has introduced its second-generation CEVA Depp Neural Network (CDNN2) this year. It has simplified machine learning in lower-power embedded systems and provided a smooth route from offline training to real-time testing with CEVA Network Generator. The machine learning algorithm can be optimized through the unique target network testing and deployment solution within a few days. The product can achieve higher efficiency with lower power loss compared to platforms that use the GPU structure. CDNN2 fully utilizes CEVA-Xm6 image forming and vision DSP processing to increase the learning speed of deep learning in embedded systems by up to four times compared to systems based on the GPU structure. At the same time, the power loss is 25 times lower with memory bandwidth achieving 15 times lower.
There are some important real world examples of smart machines. Smart home applications can use facial or voice print recognition to unlock or open doors. This can also be integrated with service suppliers. There are also examples of high-precision developments in smart farming. Supported by an integration of camera-equipped drones and Big Data analysis and prediction, farming has seen revolutionary changes. Moreover, ADAs automated systems that can integrate eight to 12 sensors combined with V2V or V2X communication standards to connect vehicles can be used to develop self-driving cars.
Always-on, always-sensing and always-connected: The three necessities of IoT devices
One of the keynote speakers, Moshe Sheier, director of strategic marketing at CEVA, showed the almost unlimited market opportunities for IoT. DSP technology promises to grow in popularity in the ever-growing and ever-changing IoT market. With the explosive development of smart mobile devices, DSP technology will play a key role to stimulate the never-ending innovative ideas in the semiconductor industry.
The continuous improvement of the semiconductor process and the design trend towards making extremely low-power loss network devices and sensors are great news for DSP technology. The smartness of today's electronic devices has greatly changed our lives. AI, as demonstrated in the cloud services offered by Amazon, Apple, and Google, shows a world of robots without a human-like exterior in voice assistant applications such as Siri and Alexa. Such technologies are expected to be widely used in other applications, triggering a revolutionary impact on people's lives.
Wearable devices, headphones, and smart medical small-size earpieces integrated with microphones, as well as Apple's AirPods, integrate the always-on technology with the extremely low-power loss technology. Through highly-integrated designs, the always-connected technology, which enables control and monitoring functions, has been seen in many smart city and smart home applications including lighting, smartwatches, vending machines and devices used by cycling teams. Self-driving vehicles integrate the always-sensing technology with other automobile technologies such as laser-based Lidar, video conferencing cameras and V2X. The IoV and machine learning technologies have been developing rapidly.
Sheier noted that today's smart IoT devices face three major challenges: more sensor integration, improvement of edge processing and new connectivity technologies to bring them all together. CEVA believes that firms should invest wisely in the always-on, always-sensing, and always-connected technologies.
Edge processing for real-time response
Once sensors have collected image and audio data to the smart device, the next challenge is edge processing. With improvements in this technology, the smart device can process video and audio data without relying on the cloud computing servers. This greatly improves the data processing ability of the devices, making it possible to respond in real-time and increase data security. This technology allows smart devices to reduce computing cost, power loss, and response time.
The key development in the drone and self-driving vehicles is edge processing, which now allows real-time control and better management of the drones through the gesture mode. Edge processing also reduces the barriers of data transmission bandwidth for applications such as drones and self-driving cars that require real-time response. With V2X data transmission technology, self-driving cars can process up to 4,000GB of data per day with edge processing. This fully utilizes the potential of DSP chips.
Although cloud computing services continue to play an important role, the technology is still controversial on the level of personal security and privacy. Smart IoT devices have local intelligence to strengthen real-time processing and this will help meet a number of requirements: security in data processing; fewer limitations on wireless transmission bandwidth; real-time response and management that is critical to personal safety; and longer battery life enabled by low power loss. The development of smart IoT devices allows DSP technology to expand to new horizons.
DSP Group voice processing chip enters new fields
DSP Group's voice recognition and voice processing chips adopt the always-on listening function, which is attractive for smart devices that aim to improve user experience. Users can give voice commands to start the timer and the step meter, and drivers can use voice control to start the GPS navigation. This technology has been adopted by South Korea-based smartphone makers.
The always-on listening technology needs to coexist with low power loss because smart devices need a long battery life. A DSP chip which uses CEVA TeakLite-410 DSP core only uses 0.2mA power during voice activation. On the voice detection mode, the chip uses less than 0.5mA of power, substantially reducing the burden on the battery.
For better consumer experience with voice control and to shorten time-to-market, DSP Group provides a full development kit for design teams to build products that supports special commands in various languages. There are also specific designs for specific phonetics used in certain Asian countries.
Voice recognition applications have been expanding. The Tokyo Olympics in 2020 is expected to have many devices that adopt voice recognition technology for announcements and interpretations. The IC design sector is aware that such applications can be easily adopted for the tourism industry. The applications can connect with the cloud services to provide interpretation services, with edge processing to help recognize the natural language. The DSP engine can demonstrate its capabilities through the always-on, noise reduction, voice control and respond, echo reduction and long-distance sound detection features.
As Sheier noted in the keynote speech, IoT can also be called "Infinite Opportunities for Technology." IoT and smart applications will be pushing the next big growth in the semiconductor and electronics industries. Through CEVA's DSP technology, IC design houses are ready for the challenge.
CEVA Director of Strategic Marketing, Moshe Sheier
Taiwan Country Manager, Barbara Yu
