Echo suppression device, echo suppression method, and echo suppression program

An echo suppression device configured to be provided in a transmitting signal path that transmits an input signal input from a microphone, in a near-end terminal including a speaker and the microphone, the echo suppression device, comprising

• a mask generation unit that
• generates a base mask based on a learning signal transmitted through the transmitting signal path when a speech is not input to the microphone and a sound is output from the speaker, and
• generates a plurality of masks by multiplying the base mask by a magnitude coefficient, the plurality of masks including the base mask;
• a mask storage unit that stores the base mask;
• a mask selection unit that sequentially causes generation of an optimal mask by the mask generation unit, the optimal mask generated for each sample point in time that a reception signal is acquired, the reception signal transmitted through a receiving signal path that transmits a signal to the speaker, the optimal mask generated according to:
• a magnitude of a reception signal and
• the base mask generated for the reception signal acquired within a predetermined period before the sample point in time;
• a double-talk detection unit that sequentially detects whether a speech is input to the microphone at the sample point in time that the reception signal is acquired based on a result of comparing the input signal with the optimal mask generated for the sample point in time;
• and
• an echo suppressor that sequentially performs a process of suppressing an echo on the input signal in response to the double-talk detection unit detecting that a speech is not input to the microphone and the reception signal includes a speech.

An echo suppression method, comprising:

• generating and storing a base mask based on a learning signal transmitted through a transmitting signal path that transmits a signal input from a microphone when a speech is not input to the microphone of a near-end terminal and a sound is output from a speaker of the near-end terminal;
• generating and storing a plurality of masks by changing a magnitude of the learning signal or by multiplying the base mask by a magnitude coefficient, the plurality of masks including the base mask;
• sequentially selecting an optimal mask from among the plurality of masks, the optimal mask selected for each sample point in time that a reception signal is acquired, the reception signal transmitted through a receiving signal path that transmits a signal to the speaker, the optimal mask selected according to:
• a magnitude of reception signal and
• the base mask generated for the reception signal acquired within a predetermined period before the sample point in time;
• sequentially detecting whether a speech is input to the microphone at the sample point in time that the reception signal is acquired based on a result of comparing the input signal with the optimal mask selected for the sample point in time;
• and
• performing echo suppression processing that suppresses an echo on the input signal in response to detecting that a speech is not input to the microphone and the reception signal includes a speech.

A non-transitory storage medium storing an echo suppression program, which is provided in a transmitting signal path that transmits a signal input from a microphone in a near-end terminal including a speaker and the microphone, causing a computer to execute:

• generating and storing a base mask based on a learning signal transmitted through a transmitting signal path that transmits a signal input from a microphone when a speech is not input to the microphone of a near-end terminal and a sound is output from a speaker of the near-end terminal;
• generating and storing a plurality of masks by changing a magnitude of the learning signal or by multiplying the base mask by a magnitude coefficient, the plurality of masks including the base mask;
• sequentially selecting an optimal mask from among the plurality of masks, the optimal mask selected for each sample point in time that a reception signal is acquired, the reception signal transmitted through a receiving signal path that transmits a signal to the speaker, the optimal mask selected according to;
• a magnitude of reception signal and
• the base mask generated for the reception signal acquired within a predetermined period before the sample point in time;
• sequentially detecting whether a speech is input to the microphone at the sample point in time that the reception signal is acquired based on a result of comparing the input signal with the optimal mask selected for the sample point in time;
• and
• performing echo suppression processing that suppresses an echo on the input signal in response to detecting that a speech is not input to the microphone and the reception signal includes a speech.