ACTS Advanced Speech Quality Analysis and Testing System

⦁ ACTS-AQA module

The voice call test includes two parts, namely the standard part and the tracking of new technologies or products part. ACTS-AQA is the standard part test, mainly used to analyze the conventional performance indicators of telecommunications and car communication audio, and meet national/international operator standards.
main standard
Telecommunications standards:
► YD/T 1538-2014
YD/T 1884-2013 (corresponding to international standards EN50332-l and EN50332-2)
CMCC Mobile Phone Testing Standards
► 3GPP2 (3GPP standard)
► 3GPP TS 26.131 /132
► 3GPP TS 26.131 /132 (SWB/FB)
► 3GPP TS 51. 010-1
► ITU-T G.165 (ITU standard)
► ITU-T G.160 Amendment 1 2012
► ITU-T G. 862
► ITU-T G. 863 POLQA
► 5-MOS/N-MOS
► Vodafone VFTST-Audio-NB
► Vodafone VFTST-Audio-WB
► Vodafone VFTST-Audio-SWB/FB
Car communication standard:
► GOST 33468-NB
► GOST 33468-WB
► ITU-T P.1100
► ITU-T P.1110
► ITU-T P.1120
► ITU-T P.1140NB
► ITU-T P.l 140WB

⦁ ACTS-MM module
An objective audio detection system that synchronizes with the development of intelligent voice technology, enabling objective testing of voice calls, human-computer interaction, and music/games on intelligent terminals.
Main Features
►Accurate reproduction of noise field
► Adjustable indoor reverberation field construction
►Provide objective standards for quality assurance and acceptance, ensuring consistency between laboratory and practical applications, thereby improving the satisfaction of intelligent terminal users
►Provide objective evaluation criteria for various noise reduction and sound pickup schemes, and provide discrimination and improvement basis for each stage of design and manufacturing
► Ensured consistency between audio and voice interaction on smart terminals
►Provide effective tools for audio performance debugging
System Features
The objective testing in the laboratory is consistent with the user experience in practical applications
Accurately evaluate performance and increase user satisfaction
Track the latest developments and applications of audio technology
Strictly reproducible
The testing performance is consistent across different laboratories and operators
► Correct comparability and clear positioning of issues
Only with strict repeatability can the performance advantages and disadvantages of various phones be strictly compared
The benefits of automated testing
► Ensure strict repeatability, minimize manual intervention, accuracy, and objectivity
► Save testing manpower and do not waste valuable talents on repetitive and error prone labor
► Improve efficiency, avoid duplicate testing, and prevent human errors

⦁ ACTS testing software
⦁ ACTS-AQA (S200) Network Call Standard Testing Software
Specially developed software for testing network call standards, mainly for testing the quality of 3GPP standard voice calls.
Basic testing items
Sending end delay
Receiver delay
Transmitting Loudness (SLR)
Receiver Loudness (RLR) - Intermediate Volume/Maximum Volume/Minimum Volume - Idle Noise at the Sending End
Idle noise at the receiving end - normal volume/minimum volume
Sending end frequency response (SFR) -1/12 octave band
Receiver frequency response (SFR) -1/12 octave band
Acoustic echo cancellation performance - normal volume/maximum volume
Double talk performance
Distorted sender
Receiver distortion

⦁ Noise playback system equalization software ACTS-BGN (S300)
S300 is a noise playback system equalization software that achieves noise playback by controlling four speakers to produce sound. It is mainly used for testing the 3QUEST (SMOS/NMOS/GMOS) module in S200 (if testing 3QUEST in S200, it is necessary to calibrate the four speakers in S300 in advance).
Main Features
Complies with the ETSI-EG 202396-1 and ETSITS 103 224 standards, supporting two calibration methods. The interface for dual ear equalization and single microphone equalization calibration is simple, intuitive, and easy to operate
The calibration results can be saved for use by S200 and S400

⦁ Multi microphone denoising system basic testing software ACTS-MM (S400)
S400 is the basic software for controlling, calibrating, configuring, and testing the ACTS-MM section. The basic testing project consists of two parts. The first part mainly focuses on the uplink testing of mobile phones, that is, after calling the phone, the artificial mouth and noise are simultaneously played to measure the external pure mixed sound without noise reduction algorithm recorded by the standard microphone and the processed sound with internal algorithm, transmission loss, and device quality recorded by the main microphone of the phone to analyze the mobile phone noise reduction algorithm. The second part includes speech recognition, voice wake-up, and speech synthesis testing for human-computer interaction. In addition, optional configuration modules are provided to expand testing capabilities.
Main Features
Includes standard near end voice and customizable or externally input near end voice bandwidth, including narrowband, broadband, and ultra wideband
Contains various standards and pre-set noise libraries (static and motion)+customizable and externally input noise libraries
4 pre-set signal-to-noise ratios+customizable signal-to-noise ratios
Reproduction of various noise fields
Basic testing items
Near end voice intensity after noise reduction Voice Level
Noise Suppression after Denoising (ITU-T P.835 Amendment 1 Objective Test)
Noise reduction performance under noise
Polar Pattern for Eliminating Noise in Various Directions
Speech recognition rate test
Voice wake-up rate test
speech synthesis

⦁ ACTS optional module
⦁ POLQA Voice Quality Measurement Module (LQT10)
POLQA is an extension and update of PESQ (ITU-T P.862), which is a testing standard for voice quality in new generation fixed, mobile, and IP networks. The new ITU-T P.863 standard, POLQA supports high-definition quality speech coding and network transmission technology, with higher accuracy for 3G, 4G/LTE, and VoIP networks.
POLQA mainly scores the speech quality in quiet and noisy environments. The higher the score, the better the speech quality. According to the standard definition, it is divided into narrowband speech and ultra wideband speech quality scores:
Narrowband voice POLQA score range: 1~4.5, ultra wideband voice POLQA score range: 1~4.75. The same narrowband voice has different POLQA scores in narrowband and ultra wideband. In narrowband score, high-frequency loss is not considered, while in ultra wideband score, high-frequency loss of narrowband signal is considered.

⦁ 3QUEST objective testing module for noisy speech quality (MOS30)
Based on the ETSI EG 202 396-3 and TS 103 106 standards, ETSI is an objective testing standard for denoised speech and residual noise, including testing modes under narrowband and broadband background noise. 3QUEST can effectively evaluate the transmission quality of speech in the presence of various background noises, obtaining three MOS scores: S-MOS, N-MOS, and G-MOS. S-MOS represents the transmission performance of speech signals, N-MOS represents the performance of background noise suppression, and G-MOS represents the overall performance.
► S-MOS: Objective Speech Quality Assessment Ignoring the Influence of Residual Noise
► N-MOS: Objective quality assessment only for residual noise
► G-MOS: Objective Speech Quality Assessment for Global Measurement of Speech and Residual Noise

⦁ Near end voice long-distance mixing environment module (SRVB10)
SRVBlO is a laboratory production system designed for measuring speech mixing environments. It mainly uses a thousand anechoic chamber and can adjust different room sizes, RT60, and wall sound absorption coefficients to produce different indoor reverberation scenes. Various speech reverberation scenarios are mainly generated by HATSl00's artificial mouth and 8 or 10 monitoring grade speakers.

⦁ ITU-G160SNRi denoising signal-to-noise ratio improvement testing module (SNRi20)
The international standard ITU-T G.760 for signal-to-noise ratio improvement testing measures the signal-to-noise ratio improvement, denoising strength, and denoising balance of denoising processing under various types of noise.
Specific parameters:
► SNRi: Improved measurement of signal-to-noise ratio after denoising, where a larger SNRi value indicates a greater improvement in signal-to-noise ratio
► TNLR: The total noise reduction dB after noise reduction, with a larger value indicating more noise reduction
► DSN: The degree to which the near end speech is reduced or amplified after noise reduction. Positive values indicate that the speech is reduced, while negative values indicate that the speech is amplified or there is too much residual noise. The closer the DSN is to 0, the better. Generally, - S<DSN<S
► DNRSL: The difference in the strength of noise suppression between the mixed and pure noise segments after denoising, with a value closer to 0 being better, generally -5<DNRSL<5.

⦁ Speech Objective Quality Improvement Evaluation Module Ignoring Residual Noise (VQi10)
Although denoising technology can reduce environmental noise, denoising algorithms may not always improve speech quality. VQi10 is used to test whether the objective quality of speech has improved or decreased compared to before and after denoising. In order to highlight speech quality, VQi10 is an objective speech quality assessment that ignores the influence of residual noise.
Applying nonlinear fractal processing theory to reverse engineering of the human ear, a new listening model is established to achieve consistent performance between objective audio detection and subjective hearing detection of the human ear, with lower distortion scores being better.

⦁ Objective Quality Assessment Module for Residual Noise (RNA20)
Fully considering factors that affect the quality of residual noise, such as musical noise (commonly known as flowing sound), it can effectively evaluate the residual noise of strong non-stationary noise such as others' speech noise after denoising processing, which is closer to the subjective evaluation of thousands of people. The score is between 1-5 points, and the higher the score, the better.

⦁ Noise reduction convergence time measurement module (CT10)
The convergence time of stationary noise is relatively easy to measure, but the convergence time of non-stationary noise is difficult to measure because the noise itself fluctuates. ACTS-CTlO overcomes this challenge by accurately measuring the convergence time of noise reduction for both stationary and non-stationary noise, measured in seconds. The shorter the convergence time, the better.

⦁ Speech intelligibility measurement module (MESTOl20)
The comprehensibility of speech is directly related to the effectiveness of human to human speech communication and the recognition rate of human to machine communication. Usually, there is reverberation and environmental noise in speech.
It will seriously affect the comprehensibility of speech. A good denoising algorithm not only preserves good speech quality, but also excellent comprehensibility. This module can effectively test the actual speech intelligibility after noise reduction processing, thereby evaluating the speech intelligibility during human to human communication and the speech recognition rate of human-computer interaction. The score ranges from 7 to 5, with higher scores being better.

⦁ Speaker Ear Reverse Engineering Distortion Testing Module (RE30)
Applying nonlinear fractal processing theory to reverse engineering of the human ear, a new listening model is established to achieve consistent performance between objective audio detection and subjective hearing detection of the human ear, with lower distortion scores being better.

⦁ Active noise reduction performance measurement module for downlink earpiece (ANC10)
Active noise reduction at the receiving end - a method to solve the problem of making the user hear the other party's voice clearly under environmental noise. ACTS-ANClO is used to objectively test the performance of active noise reduction at the receiving end. Measure the noise frequency response under various grip positions, pressures, and states (call/standby) to verify the actual effect of active noise reduction.
► ANC vs Talking/Idle
Measure the frequency response curves of the specified pressure and distance from the head during mobile phone calls and standby, respectively, to obtain complete ANC spectrum characteristics.
► ANC vs Pressure
Measure the noise frequency response curve for several consecutive pressure values and away from the human head during a call to obtain the differences and trends in ANC performance under different pressure conditions.