基于语音识别技术的残疾人环境控制系统的研制

The Development of a Speech Recognition Based Environmental Control Unit For the Disabled

本文介绍了一种基于语音识别技术的残疾人环境控制系统 (ECU) ,该系统用 PC机采集和识别输入语音 ,通过并口将指令传到控制箱中。控制箱内主要是译码电路、继电器组和电视红外遥控电路 ,分别控制电器设备、电话机和电视机的各项功能。该系统采用小词汇表孤立词语音识别模式 ,在语音指令的识别中用短时能频值作为切分音节的依据 ,用 L PC倒谱和差分倒谱作为特征矢量 ,用倒谱域非线性分块法对特征矢量进行压缩和时间对准 ,最后用最近邻法作为判别准则。在识别算法中采用字长阈值法、滑动模板法和误差双阈值法来减小环境噪声的影响。经过测试 ,该系统对于指令词的识别率高 ,对环境噪声不敏感 ,硬件的控制可靠 。

A Speech Recognition Based Environmental Control Unit(ECU) is presented in this paper. In the system, a PC is used to sample the input speech and recognize it. Recognized results, which are the control commands, are sent to the control box through a parallel port of the computer. In the control box there are a decoding circuit, groups of relays, and a circuit of infrared remote controller. The control box controls electric appliances, a telephone, and a TV set for some functions. In the recognition of commands, which is speaker dependent small vocabulary discontinuous speech recognition, Energy Frequency Value threshold is used to separate syllables. Coefficients of Linear Predictive Coding(LPC) Cepstrum and delta Cepstrum are selected as eigenvectors, which are compressed and time registered according to Nonlinear Blocking rule in the Cepstrum field. The nearest neighborhood criterion is used. Threshold of word length, sliding template, and double threshold of error are applied to minimize influence of environmental noise. Test result shows that this ECU has a high recognition rate for commands; it is insensitive to environmental noise, and the hardware control is reliable. This ECU is suitable to be used for high level paraplegic users.