Design and Implementation of a SoPC System for Speech Recognition
This paper presents the design of a System on Programmable Chip (SoPC) based on Field Programmable Gate Array (FPGA) for speech recognition in which Mel-Frequency Cepstral Coefficients (MFCC) for speech feature extraction and Vector Quantization (VQ) for
- PDF / 1,746,766 Bytes
- 7 Pages / 439.37 x 666.142 pts Page_size
- 2 Downloads / 232 Views
Abstract This paper presents the design of a System on Programmable Chip (SoPC) based on Field Programmable Gate Array (FPGA) for speech recognition in which Mel-Frequency Cepstral Coefficients (MFCC) for speech feature extraction and Vector Quantization (VQ) for recognition are used. The execution speed of the blocks in the speech recognition system is surveyed by calculating the number of clock cycles while executing each block. Keywords Speech recognition
MFCC VQ SoPC FPGA Nios
1 Introduction Speech recognition system is applied in many application fields such as health care, military, human computer interaction, avionics technicians… [1], especially, the applications which support disabled people to communicate with the world in a better way. For that reason, there are many studies on software/hardware implementation of speech recognition systems for many years. However, because of a large number of accents spoken around the world, there are still many challenges that need further research and development, for example, Vietnamese speech recognition. T. Van Hoang N. L. T. Truong H. Trang (&) University of Technology, Vietnam National University, HoChiMinh City, Vietnam e-mail: [email protected] T. Van Hoang e-mail: [email protected] N. L. T. Truong e-mail: [email protected] X.-T. Tran VNU University of Engineering and Technology, 144 Xuan Thuy, Hanoi, Vietnam e-mail: [email protected]
J. J. (Jong Hyuk) Park et al. (eds.), Multimedia and Ubiquitous Engineering, Lecture Notes in Electrical Engineering 240, DOI: 10.1007/978-94-007-6738-6_147, Springer Science+Business Media Dordrecht(Outside the USA) 2013
1197
1198
T. Van Hoang et al.
The research on speech recognition going mainly in two directions, namely: the software runs on Personal Computers (PCs) and embedded systems. For the first direction, many studies and software tools have been developed successfully. In particular, the Hidden Markov Model Toolkit (HTK) is a toolkit for building Hidden Markov Models (HMMs) used in speech recognition successfully [2]. There are also many tools running on the PC or smart phone aimed at the control device via speech. For the second direction, embedded systems have many advantages as high performance, convenience, low cost, and great development potential. However, speech recognition research based on embedded systems is more difficult. This paper will present the implementation of a speech recognition system as an embedded system using FPGA technology. In fact, the implementation of speech recognition systems has been done using FPGA technology in recent years. In paper [3], speech recognition systems are implemented as hardware/software co-design systems using Hidden Markov Model (HMM). This project use Linear Predictive Coding (LPC) method in feature extraction block. So, the recognition accuracy is not high compared with the MFCC method. In paper [4], the MFCC method is applied, but the optimization was not taken into account yet to increase performance. Another work, presented in [5
Data Loading...
