Research And Implementation Of Speech Recogniion Technology In Automatic Outbound Call Centers

Nowadays, call center service has been widely used in various industries, including service class call center and marketing class call center. They’re developing in an astonishing speed, especially the latter ones. After all, Call Center industry is experiencing a significant moment.Speech recognition technology has significant impact on daily life. Some related products have been applied to the telephone network, computer networks, mobile terminals, banking self-service, medical devices and industrial devices. Yet how to integrate Speech recognition technology with call center is still a challenge.The automatic outbound call of call center means that an enterprise has a task list of outbound call, and has the call center to complete the outbound call task instead of agent. According to the agent’s status and corresponding arithmetic strategy, call center makes automated calls to those assigned number list. The probably solutions call center will make after analyzing numbers are as follows:accessing to agent if the phone is answered, calling next number and recording the call statuses automatically if the phone is not answered.Since the volume of outbound call business is doubled nowadays, efficiency means sales volume and profit in outbound call marketing. The number identification system can filter numbers, so as to reduce invalid outbound calls from call center evidently, improve efficiency of agents, and then create profits accordingly. Therefore in the application of outbound call center, both proactive dial and predictive dial have abundant demand for number identification. As a result, it’s been urgent to figure out how to filter outbound call numbers accurately, how to improve dial efficiency to help agents spend more time on the interaction with customs rather than just wait.This paper presents a novel automatic outbound calls system utilizing speech recognition technology. Combining signalling inspection and tone inspection, its accuracy can reach 90%. Moreover, compared with traditional systems, the dialing efficiency has greatly improved and the total cost of call center has also been reduced.The core of the speech recognition module is to divide the spectrum of the speech signal into several sub-bands. Tones can be identified through analyzing the maximum energy value of each sub-band, comparing the length of mute segments and the effective voice segments with standard ones. By analyzing the features of live Speech and using a series of heuristic rules can distinguish them from unnatural sounds. Finally, by categorizing information of the database, then choosing proper distance and searching methods, AM can be quikly found.In practice, the system has achieved pretty good results in improving the efficiency of cooperative enterprise. I believe it will have a bright future in the call center industry.