Directional Activity Mask Detector with Blocking Matrix for Vehicle Speech
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech enhancement systems in vehicles face challenges in accurately detecting the target speaker's activity due to imprecise activity detection, leading to self-cancellation and interference leakage, especially when using Relative Transfer Function Tracking (RTFT) with incorrect direction updates.
Innovation Solution
A directional activity mask detector using an algebraic blocking matrix is applied to a microphone array, generating a blocking matrix based on pre-recorded signals, detecting unblocked zones, and estimating a Relative Transfer Function (RTF) vector to enhance target signals while ignoring directional interference.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If RTFT is updated with wrong direction, then direction tracking may be contaminated, but self-cancellation occurs and target signal quality deteriorates
Solution Approach 1:
The blocking matrix is pre-computed based on pre-recorded signals from different zones before actual speech enhancement operation. This preliminary preparation allows the system to quickly apply appropriate blocking without real-time computation delays, preventing both self-cancellation and interference leakage by having the correct blocking configuration ready in advance.
Solution Approach 2:
The blocking matrix acts as an intermediary component between the microphone array input and the RTFT estimation process. It selectively blocks signals from specific zones before they reach the RTFT tracker, preventing contamination of direction tracking while allowing the target speaker's signal to pass through unblocked zones to the beamformer.
2Measurement precision
If activity detection is imprecise, then interference leakage into target output increases, but improving detection accuracy increases system complexity
Solution Approach 1:
The spatial environment is segmented into multiple zones around the vehicle, with each zone having its own pre-computed blocking configuration. The blocking matrix is constructed by selecting and combining zone-specific blocking vectors based on the detected active zones. This segmentation approach simplifies activity detection by reducing it to zone-level identification rather than full-spectrum analysis.
Solution Approach 2:
Different blocking configurations are prepared for different spatial zones, allowing the system to apply locally optimized blocking strategies. Each zone has its own pre-recorded signals and corresponding blocking matrix columns, enabling the system to handle each zone's acoustic characteristics independently while maintaining overall system simplicity.
3Object-affected harmful factors
If blocking matrix is applied to block target zone, then interference from other zones is reduced, but target signal may be blocked causing loss of useful information
Solution Approach 1:
The blocking matrix application is dynamically controlled based on real-time zone activity detection. The system selectively applies blocking only to zones that are detected as inactive or containing interference, while leaving blocks on zones identified as containing the target speaker. This dynamic adaptation prevents target signal loss while maintaining interference reduction benefits.
Solution Approach 2:
The blocking matrix parameters (which columns are applied and to which zones) are changed based on the detected active zones and target speaker location. The system adjusts the blocking configuration by selecting different columns from the pre-computed blocking matrix corresponding to different spatial zones, optimizing the balance between interference rejection and target signal preservation.
Data Source
AI summary
A method for a directional activity mask detector for a vehicle includes generating a blocking matrix based on pre-recorded signals from a target zone, receiving, at a voice activity detector, audio frames from a microphone array, and applying the blocking matrix to one or more zones within a vehicle. The method also includes detecting signals from unblocked zones of the vehicle, determining an activity of a target signal based on the detected signals from the unblocked zones, and estimating, by a beamformer, a relative transfer function (RTF) vector based on the received audio frames and the determined activity of the target signal.