Development of Multi-Modal Biometric Authentication Chip toward Ubiquitous Network Society

1.引入笔移角度、加速度和矢量参数，并将其融合到其他参数中，提高了小波域在线签名验证的性能。此外，还发现最优阈值的变化与签名的复杂性有关。根据签名复杂度的不同，提出了门限均衡方法。结果达到了99.0%的验证率。提出了利用基音进行语音验证的方法，并对其性能进行了测试。验证率为74.3%。此外，还对利用倒谱进行的语音验证进行了检验。验证性能提高到88.5%。同时，提出了范数约简方法。研究了融合在线签名和语音的多模式生物特征认证。采用加权求和法得到了97.5%的验证率。此外，还证实了调整每个模型中的阈值可以提高验证率。我们尝试在基于Windows Mobile操作系统和Visual Studio 2005.5的掌上电脑系统中实现多模式生物特征认证。我们尝试在现场可编程门阵列中利用在线签名和语音实现多模式生物特征认证，并在现场可编程门阵列中实现了基音检测中的自相关功能。

1. Pen-movement angle, acceleration and vector parameters were introduced and fused in other to improve the performance of DWT domain on-line signature verification. In addition, it was found that the variation of optimum threshold values had relation with the complexity of signatures. We proposed threshold equalizing methods based on the complexity of the signature. As a result, the verification rate of 99.0% was achieved.2. Voice verification using the pitch was proposed and its performance was examined. The verification rate was 74.3%. Additionally, the voice verification using the cepstrum was examined. The verification performance was improved to 88.5%. Also, the norm reduction method was proposed.3. Multi-modal biometric authentication fusing the on-line signature and the voice was examined. The verification rate of 97.5% was achieved using the weighted sum rule. Also, it was confirmed that adjusting the threshold value in each modal improved the verification rate.4. We tried to implement the multi-modal biometric authentication in the PDA system based on Windows Mobile OS and Visual Studio 2005.5. We tried to implement the multi-modal biometric authentication using the on-line signature and the voice in FPGA.In this time, the auto-correlation function in pitch detection was implemented in FPGA.