Coprocessor as an independent hardware accelerator for improved user experience

Today's consumers are increasingly demanding technology, which is particularly evident in the interface technology used to interact with devices. People are increasingly demanding devices, including mobile phones, in-vehicle electronics, home networking, and office environments, requiring them to incorporate an easier-to-use, more intuitive user interface to better reflect the interactions between people.

In every new product cycle, designers are under pressure to design products that are more accurate and intuitive. In recent years, after touch-screen technology has gradually spread to include devices such as phones, tablets, monitors, point-of-sale solutions, ATMs, and inquiry machines, voice recognition technology is rapidly becoming the next-generation user interface technology that drives product innovation and application. Speech recognition, even gesture and image recognition, is a standard issue for all kinds of work and personal equipment. Since speech recognition technology is subject to the development of embedded applications to a certain extent, it is still in the early stage of development. However, the speech recognition interactive interface technology will eventually be widely adopted, which is the general trend of technology development. The automotive industry is already planning to introduce more sophisticated modules with enhanced embedded speech recognition capabilities.

The slow development of speech recognition technology is partly due to the more intuitive user interface, the exponential increase in processing power and memory required, which in turn promotes the innovative development of flash memory technology. As most designers know, the more intuitive the user interface, the more complex the technology platform and design it requires. User interface technology will consume more computing power and flash memory to achieve the best user experience while achieving high performance processing power. One solution is to use specialized hardware, a dedicated coprocessor with next-generation flash capabilities and integrated logic and flexible software algorithms. These coprocessors can be used as a stand-alone hardware accelerator to offload the main application processor to achieve the highest level of user experience on the market.

Evolution of human-computer interface

Since the advent of computer mice, HMI (Human Machine Interaction) technology has made great strides. User interface innovations have historically been attributed to the successful use of new devices, for example, from physical buttons on older mobile phones to touch screens on smartphones. Creating an engaging user interface is challenging and requires a fairly complex system to create a functional, easy-to-access, logically clear and enjoyable user experience. This complex system requires high reliability and high performance hardware in terms of processing power and flash bandwidth. As the core functional innovations of end products are nearing maturity, consumers are increasingly making purchase decisions by using the industrial design and user interface of the product as a standard. Manufacturers have also noticed this change, and flash manufacturers and designers are also under market pressure to accelerate innovation in response to market demands. Speech recognition is the focus of the next wave of human-computer interaction technology innovation.

How does speech recognition work?

While advanced HMI technology is increasingly becoming the de facto standard for many consumer electronics products, high-performance processing capabilities are becoming more critical for embedded systems. In general, speech recognition can be broken down into three processing stages:

The first phase is the sound processing phase, which typically takes less than 5% of the processing power, ie the system converts the captured sound signals from analog information to digital information. This is also the stage of filtering, suppressing noise, and echo cancellation, distinguishing the microphone sound from the falsely captured noise. The processed signal is output in the form of a digital sound stream, and each segment of the sound is unique like a fingerprint. The second phase is the matching phase, where the system matches these sound signals with the "speech library", the acoustic model. This matching phase, called acoustic scoring, takes up 50% to 70% of the system's processing bandwidth. The acoustic scores generated in the second stage will enter the third stage as input information, that is, the system translates these acoustic signals into text information by searching the language and the dictionary model. This phase will take up 30% to 50% of the processing power.

In general, the entire process is handled by one CPU, and this processor is also responsible for handling several other tasks. Because speech recognition is very computationally intensive and flash space, sharing resources in an embedded solution can lead to unacceptable delays or limit the ability of bandwidth to handle increasingly software models. In order to achieve greater accuracy, the size of the software model is expanding.

Why are there special hardware for the HMI process?

Due to the heavy storage and computational bandwidth limitations of HMI (eg, speech recognition) processing, this multitasking method of sharing one CPU resource is often at the expense of some end user experience.

For example, in speech recognition, designers must choose between speed and accuracy under embedded system conditions that share resources. Larger acoustic models can achieve higher accuracy, but with greater processing power to avoid unacceptable delay response. In addition, as users increase their expectations for voice processing interfaces, such as the desire to distinguish between gender, noise, dialogue, accent, and multi-language, the size of this feature-rich voice module is exponentially increasing, and reliability. High, fast-access memory will become even more important for this ever-increasing performance. Unfortunately, today's resource-sharing, resource-constrained hardware platforms do not provide acceptable processing power for today's largest acoustic models. As a result, the industry is currently only second to none, developing a compressed version of the acoustic model that provides the lowest acceptable accuracy in a minimum acceptable response time.

To overcome this shortcoming, the industry has recently developed a solution: a dedicated hardware coprocessor that boosts processing power and speeds up certain speech recognition processing stages. The first representative of this type of solution is the Spansion voice coprocessor. The Spansion Voice Coprocessor is responsible for the acoustic scoring phase of speech recognition, thereby sharing the CPU load and reducing response latency by up to 50%. In addition, the Spansion Voice Coprocessor provides enough processing power for today's largest acoustic models up to 10 times the current common acoustic model. This solution can completely solve the trade-off between delay and accuracy, which is a major design problem in embedded speech recognition solutions based on resource sharing platforms.

These specialized user interface coprocessors utilize advanced flash technology to achieve instant response, high reliability and high performance for a range of applications.

What does an advanced HMI look like?

Although speech recognition represents the current trend of embedded system HMI development, at present, we have only taken a small step in providing advanced HMI for natural language understanding, image recognition or emotion perception. However, the development of flash memory technology in recent years is constantly pushing the entire industry to a newer and more creative development level. Dedicated hardware combined with advanced flash and logic devices provides the foundation for a more powerful and powerful software model that will lead us closer to the future of more sophisticated HMI technology. As the pace of flash innovation continues to advance, we will be better able to design a richer, more natural interface that ultimately improves the user experience.

Consumers' pursuit of the best user experience continues to spur us to create new structures. Therefore, it is users who are constantly pushing the continuous innovation of flash memory technology and moving forward rapidly. Today, the forefront and challenge of the next technology development is to provide a richer user experience, with more powerful voice recognition capabilities with advanced flash technology and dedicated hardware.

Car Inverter

Car Inverter,150W Car Inverter,Dc Ac Car Converter,Inverter For House Car

GuangZhou HanFong New Energy Technology Co. , Ltd. , https://www.hfsolarenergy.com