Speech Enhancement Technology And DSP System Design In Intelligent Conference Environment

With the rapid development of information technology,the demand for intelligent conference rooms among corporate units has gradually increased.The audio equipment occupies an important role in the intelligent conference room and is widely used in information technology such as on-site conference equipment,remote video conference equipment,and remote teaching equipment.In fact,voice signals are unavoidably interfered with by various types of noise in the process of collection or transmission,resulting in the decrease of subjective hearing quality.Speech enhancement technology is a method of noise reduction processing.It can extract useful information in noisy environment,reduce noise,and improve speech quality and intelligibility.The listeners are willing to accept and do not feel tired.It can play a good effect in the intelligent conference system.With the background of the intelligent conference room,this paper develops and implements speech enhancement system of intelligent voice control under the project requirements of a technology company in Henan.This paper first introduced the basic knowledge of speech enhancement,and then studied several methods of speech enhancement,including power spectral subtraction,wiener filtering and minimum mean square error algorithm.The MATLAB simulation tests were conducted in different noise environments,and enhancement effect of each algorithm was analyzed.Among them,power spectral subtraction is simple and efficient,and the amount of operation is small.It becomes a common method for speech enhancement.The disadvantages are that the noise estimation error is too large and the music noise residual problem.Based on the shortcomings of the traditional power spectral subtraction,this paper proposed an improved power spectral subtraction algorithm based on subspace decomposition.Subspace decomposition is used to obtain the eigenvalue to realize the noise energy estimation in noisy speech,and the more accurate spectral subtraction is performed.Then,the audio data frames are evaluated by the signal energy after spectral subtraction,and the speech frames and noise frames are effectively separated.Finally,the speech frames are smoothed to suppress the music noise,and the noise frames are attenuated to further reduce the noise residual.Experiments on noisy speech enhancement processing and evaluation of parameter measurements show that the proposed algorithm can effectively estimate the noise energy,and the noise residual degree is smaller and the performance is better.The hardware design of speech enhancement system controlled by the intelligent voice is divided into two subsystems: the sending board and the receiving board.The sending board uses the Bluetooth chip CSR8670 for voice wireless transmission,and the receiving board adopts DSP chip TMS320VC5509 A with high-speed frame operation in order to ensure the real-time performance of voice processing.The hardware circuit prototype was designed,and the hardware and software were jointly adjusted,and then the improved power spectrum reduction enhancement algorithm was downloaded to the hardware platform to complete the design of the entire speech enhancement system.After testing and verifying,the system can effectively suppress noise interference in real time,and the quality of the processed speech is improved,which achieves the requirements of the control speech enhancement in the actual intelligent conference.