Speech recognition technology has been a very active research field in recent years, and the practical research of speech recognition systems is a major direction of speech recognition research. In recent years, the demand for low-cost and highly robust speech recognition functions in consumer electronic products has increased rapidly, and speech recognition systems have been transferred from laboratory PC platforms to embedded devices in large quantities. Embedded devices are usually designed for specific applications and only need to recognize commands of dozens of words, which belongs to small vocabulary speech recognition systems. Therefore, the requirements for speech recognition technology are not large vocabulary and continuous speech recognition, but recognition accuracy and robustness. For embedded systems, there are many other factors to consider. The first is cost. Generally, fixed-point DSP is used, and sometimes only MCU can be considered, which means that the complexity of the algorithm is limited; secondly, embedded systems have strict restrictions on volume, which requires a highly integrated hardware platform. Therefore, embedded systems with SOC structure greatly reduce the number of chips, can provide highly integrated and relatively low-cost solutions, and also greatly improve the reliability of the system. Speech recognition system-on-chip is a system-level integrated chip. It not only puts several complex digital logic circuits into the same chip to make a complete single-chip digital system, but also integrates other functional devices such as AD/DA and memory in the chip. The author uses the SPCE061A chip of Lingyang Company to realize a stable and reliable embedded non-specific person speech recognition system.
You Might Like
Recommended ContentMore
Open source project More
Popular Components
Searched by Users
Just Take a LookMore
Trending Downloads
Trending ArticlesMore