Abnormal Condition Detection Using Voice and Acceleration Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques for detecting abnormal vehicle conditions, such as accidents, require pre-prepared scream data and are inadequate for recognizing unprepared screams, limiting their effectiveness in real-time detection.
Innovation Solution
An abnormal condition detection system that includes a voice abnormality detector generating a normal time model from voice data and a vehicle body abnormality detector using acceleration data, with an abnormal condition determiner combining results to assess vehicle status without pre-prepared scream data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If scream data is prepared in advance using conventional techniques, then the system can recognize prepared scream types, but it cannot recognize unprepared screams and requires large amounts of pre-collected data
Solution Approach 1:
Instead of preparing scream data in advance and matching incoming sounds against it, the patent inverts the approach by collecting normal voice data in advance, building a voice profile from it, and then detecting deviations from this profile. This allows the system to recognize any abnormal sound including unprepared scream types without needing pre-collected scream samples.
Solution Approach 2:
The patent changes the detection parameter from matching specific scream characteristics to detecting deviations in voice parameters from the normal profile. By monitoring changes in voice parameters against the established baseline, the system can identify any abnormal vocalization regardless of whether it was pre-programmed.
2Adaptability or versatility
If multiple types of scream data are collected in advance, then more scream variations can be recognized, but the data volume becomes excessively large
Solution Approach 1:
The patent extracts only the essential voice characteristics from normal speech to create a compact voice profile, rather than storing large amounts of actual scream samples. This extracted profile contains the necessary information to detect abnormalities without requiring the extensive data storage of multiple scream variations.
Solution Approach 2:
Instead of storing actual scream recordings, the patent creates a simplified copy or representation of the voice characteristics through the voice profile. This copy captures the essential patterns needed for detection while occupying minimal storage space compared to storing diverse scream samples.
3Measurement precision
If conventional scream detection methods are used, then prepared scream models can be identified, but the system lacks effectiveness for real-time detection of unprepared abnormal conditions
Solution Approach 1:
The patent implements a dynamic detection system where the voice profile serves as a flexible baseline that can adapt to detect any deviation. Unlike static scream models that only recognize pre-programmed types, this dynamic approach enables real-time detection of any abnormal vocalization pattern, including previously unseen scream types.
Data Source
AI summary
An abnormal condition detection system includes a voice abnormality detector that includes a normal time model generator for generating a model of normal time voice data of a person riding in a vehicle as a normal time model based on voice data including voice of the person, and detects an abnormality in the voice based on a current voice data of the person and the normal time model; a vehicle body abnormality detector that detects an abnormality in a vehicle body of the vehicle based on acceleration data of the vehicle; and an abnormal condition determiner that determines whether or not the vehicle is in an abnormal condition based on a detection result of the voice abnormality detector and a detection result of the vehicle body abnormality detector.