Scholarlink

<p>Speech recognition serves as a crucial human-computer interaction interface in intelligent systems, with broad applications in smart homes, autonomous vehicles, and companion robots. However, existing Chinese speech recognition solutions often suffer from high deployment costs, strong dependence on personal computers, limited performance, and complex system maintenance, which significantly hinder their adoption in portable and embedded scenarios. To address these challenges, this paper proposes a low-cost, portable, and high-performance Chinese speech recognition system based on the integration of Raspberry Pi and Baidu AI Cloud services. The system incorporates a spectral subtraction algorithm to suppress the impact of environmental noise on recognition accuracy. We evaluated the proposed approach on public benchmark datasets. Experimental results demonstrate that the developed system achieves a Character Error Rate (CER) of 5.85% in quiet environments. Under strong stationary noise conditions with a signal-to-noise ratio (SNR) of 0 dB, the CER is 15.6%. In the presence of strong non-stationary noise, the system maintains comparable recognition performance to the baseline. In summary, the proposed system exhibits competitive recognition accuracy in clean and stationary noise environments, and offers significant advantages in terms of cost, portability, and ease of deployment, highlighting its great potential for embedded and mobile intelligent devices.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Chinese speech recognition system based on raspberry Pi

  • 期刊论文

  • 发表日期：

    2026年3月5日

查看全文
  • Xiaopeng Wang,
  • Guoliang Zhou,
  • Zhengyang Tang,
  • Bo Xu,
  • Hongpeng Hua,
  • Fanwei Meng

摘要

Speech recognition serves as a crucial human-computer interaction interface in intelligent systems, with broad applications in smart homes, autonomous vehicles, and companion robots. However, existing Chinese speech recognition solutions often suffer from high deployment costs, strong dependence on personal computers, limited performance, and complex system maintenance, which significantly hinder their adoption in portable and embedded scenarios. To address these challenges, this paper proposes a low-cost, portable, and high-performance Chinese speech recognition system based on the integration of Raspberry Pi and Baidu AI Cloud services. The system incorporates a spectral subtraction algorithm to suppress the impact of environmental noise on recognition accuracy. We evaluated the proposed approach on public benchmark datasets. Experimental results demonstrate that the developed system achieves a Character Error Rate (CER) of 5.85% in quiet environments. Under strong stationary noise conditions with a signal-to-noise ratio (SNR) of 0 dB, the CER is 15.6%. In the presence of strong non-stationary noise, the system maintains comparable recognition performance to the baseline. In summary, the proposed system exhibits competitive recognition accuracy in clean and stationary noise environments, and offers significant advantages in terms of cost, portability, and ease of deployment, highlighting its great potential for embedded and mobile intelligent devices.