Live Call Detection in Automated Telephony Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated calling systems face inefficiencies due to delayed detection of live calls versus automated responses, leading to misidentification and nuisance calls, which are regulated due to high false positive rates and inefficient agent time usage.
Innovation Solution
A live call detection system using a telephony card and server computer algorithm that analyzes voice signals and message interactions to differentiate between human and automated responses through pre-recorded messages and voice synthesis, ensuring seamless transition to live agents within acceptable delays.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If automatic answering machine detection is implemented using traditional voice energy analysis, then calls answered by machines can be identified and routed away from agents, but the detection process takes several seconds causing live callers to experience unacceptable delays and potential hang-ups
Solution Approach 1:
The system performs preliminary actions by immediately playing a detection message when a call is answered, rather than waiting for voice energy patterns to develop. This allows the detection process to begin instantly, analyzing the recipient's response to the detection message within a predetermined time period, thereby reducing the delay experienced by live callers while maintaining detection accuracy.
2Loss of time
If the detection time is limited to a maximum of two seconds to meet acceptable call delivery standards, then live call delivery is improved, but the confidence level in determining whether the call is answered by a machine or person decreases, resulting in false positives
Solution Approach 1:
The system uses feedback by analyzing the recipient's response to a detection message that is played immediately upon call answer. The system evaluates whether the recipient spoke during the predetermined time period (typically 2-5 seconds), providing clear binary feedback (spoke or didn't speak) that enables rapid determination with high confidence. This feedback mechanism allows accurate classification within the 2-second window required for acceptable call delivery, eliminating false positives by using the presence or absence of speech response as the determining criterion.
3Reliability
If traditional voice energy analysis with pause detection is used, then answering machines can be detected based on their recorded message patterns, but the system generates false positives where live callers are misidentified as machines, causing anxiety and regulatory concerns
Solution Approach 1:
The system introduces an intermediary detection message as a mediator between the automated calling system and the recipient. This detection message serves as a standardized probe that elicits a consistent type of response (or lack thereof) from both live callers and answering machines. By using this intermediary, the system creates a controlled interaction where the absence of speech during the predetermined time period reliably indicates an answering machine, while the presence of speech indicates a live caller, thereby eliminating false positives and the associated caller anxiety.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method of operating a telephony system for determining whether a telephone call made to a call destination terminal is a live call which is answered by a live person and not by an answering machine, the system performing steps comprising detecting that the telephone call has been answered and that the call is in progress transmitting one or more voice messages in a transmitted signal to the call destination terminal analysing a received signal from the call destination terminal to detect voice signals, and determining whether or not the telephone call is a live call by applying test criteria to relative timing of events in the received signal and transmitted signal, the test criteria being based on an expected pattern of speech and silence representative of normal conversation, wherein the call is determined to be a live call according to the test criteria if, one of the voice messages is an uninterrupted voice message during which no voice signal in the received signal is detected, and a voice signal is then detected in the received signal within a predetermined listening period, P2, after the uninterrupted voice message has ended.