SSTAR SE ALGORITHM USER GUIDE

---

REVISION HISTORY

Revision No.	Description	Date
1.1.0	Initial release	05/01/2021
1.1.1	Added noise type and intensity option Added IaaSe_Set function	06/25/2021
1.1.2	Modified IaaSe_Run API and library name	06/30/2021
1.1.3	Added stereo process mode	07/01/2021
1.1.4	Added IaaSe_GetInputSamples API Added more detail description	07/12/2021
1.1.5	Modified Iaa_Run API and example	08/12/2021
1.1.6	Modified error code	10/18/2021
2.0.0	2.0 version	11/29/2022
2.0.1	Modified API version	04/11/2022
2.1.0	update normalize	12/14/2023

---

1. OVERVIEW

1.1. Algorithm Introduction

The Speech Enhancement (SE) algorithm uses AI algorithms to enhance input speech. It can be used to suppress both stationary noise and non-stationary noise.

1.2. Algorithm Specification

The working sampling rate of this algorithm is 8k/16kHz, and the length of each frame is 128 sampling points (8ms).

---

2. API INTRODUCTION

2.1. Function Module API

API Name	Function
IaaSe_GetBufferSize	Get the memory size required to run SE algorithm.
IaaSe_Init	Initialize SE algorithm.
IaaSe_SetConfig	Set SE algorithm parameters.
IaaSe_GetConfig	Print out SE algorithm parameters.
IaaSe_GetInputSamples	Get the number of sampling points input by SE algorithm.
IaaSe_Run	Run SE algorithm.
IaaSe_Free	Release resource from SE algorithm.
IaaSe_setCallbackFunc	Callback function of SE algorithm verification.

2.2. IaaSe_GetBufferSize

• Function

  Get the memory size required to run SE algorithm.

• Syntax

  int IaaSe_GetBufferSize(void);
  

• Parameter

  Parameter Name	Description	Input/Output

• Return Value

  The return value is the memory size required to run SE algorithm.

• Dependency

  • Header File: AudioSeProcess.h

  • Library File: libSE_LINUX.so/ libSE_LINUX.a

• Note

  This interface only returns the memory size required to run SE algorithm; to apply and release the memory, you need to use other APIs.

• Example

  N/A.

2.3. IaaSe_Init

• Function

  Initialize the memory required to run SE algorithm.

• Syntax

  SE_HANDLE IaaSe_Init(char* workBufAddress, AudioSeInit_t *seInit);
  

• Parameter

  Parameter Name	Description	Input/Output
  workBufAddress	The memory address used by SE algorithm.	Input
  seInit	Pointer to the initialization structure of SE algorithm.	Input

• Return Value

  Return Value	Result
  handle	Successful.
  NULL	Failed.

• Dependency

  • Header File: AudioSeProcess.h

  • Library File: libSE_LINUX.so/ libSE_LINUX.a

• Note

  • The SE algorithm only supports 16kHz sampling rate and 16bit sampling width.

• Example

  N/A.

2.4. IaaSe_SetConfig

• Function

  Set parameters of SE algorithm.

• Syntax

  int IaaSe_SetConfig(SE_HANDLE handle, AudioSeConfig_t seConfig);
  

• Parameter

  Parameter Name	Description	Input/Output
  handle	The handle of SE algorithm.	Input
  seConfig	Pointer to the parameter configuration structure of SE algorithm.	Input

• Return Value

  Return Value	Result
  0	Successful.
  Non-Zero	Failed. Please refer to Error Code

• Dependency

  • Header File: AudioSeProcess.h

  • Library File: libSE_LINUX.so/ libSE_LINUX.a

2.5. IaaSe_GetConfig

• Function

  Set parameters of SE algorithm.

• Syntax

  int IaaSe_GetConfig(SE_HANDLE handle);
  

• Parameter

  Parameter Name	Description	Input/Output
  handle	The handle of SE algorithm.	Input

• Return Value

  Return Value	Result
  0	Successful.
  Non-Zero	Failed. Please refer to Error Code

• Dependency

  • Header File: AudioSeProcess.h

  • Library File: libSE_LINUX.so/ libSE_LINUX.a

2.6. IaaSe_GetInputSamples

• Function

  Get the number of the sampling points input by SE algorithm.

• Syntax

  int IaaSe_GetInputSamples(SE_HANDLE handle, int *samples);
  

• Parameter

  Parameter Name	Description	Input/Output
  handle	The handle of SE algorithm.	Input
  samples	The number of the sampling points input by SE algorithm.	Output

• Return Value

  Return Value	Result
  0	Successful.
  Non-Zero	Failed. Please refer to Error Code

• Dependency

  • Header File: AudioSeProcess.h

  • Library File: libSE_LINUX.so/ libSE_LINUX.a

• Note

  • This function is optional. The return value of the SE algorithm is 128 for single channel and 256 for dual channels. According to the sampling bitwidth of 16 bits, the memory size of the single-channel input array is 128 * 16 bit, and the memory size of the dual-channel input array is 256 * 16 bit.

2.7. IaaSe_Run

• Function

  Run SE algorithm.

• Syntax

  int IaaSe_Run(SE_HANDLE handle, short *input);
  

• Parameter

  Parameter Name	Description	Input/Output
  handle	The handle of SE algorithm.	Input
  input	The pointer to input data.	Input/Output

• Return Value

  Return Value	Result
  0	Successful.
  Non-Zero	Failed. Please refer to Error Code

• Dependency

  • Header File: AudioSeProcess.h

  • Library File: libSE_LINUX.so/ libSE_LINUX.a

• Example

  Please refer to Demo code.

2.8. IaaSe_Free

• Function

  Release the resource from SE algorithm.

• Syntax

  int IaaSe_Free(SE_HANDLE handle);
  

• Parameter

  Parameter Name	Description	Input/Output
  handle	The handle of SE algorithm.	Input

• Return Value

  Return Value	Result
  0	Successful.
  Non-Zero	Failed. Please refer to Error Code

• Dependency

  • Header File: AudioSeProcess.h

  • Library File: libSE_LINUX.so/ libSE_LINUX.a

• Note

  • Please call IaaSe_Free before releasing the memory used by SE algorithm.

2.9. IaaSe_setCallbackFunc

• Function

  The callback function of SE algorithm verification (currently not supported).

• Syntax

  int IaaSe_setCallbackFunc(int(*log)(const char *szFmt, ...),
                   int(*envSet)(char *key, char *par),
                   int(*envGetString)(char *var, char *buf, unsigned int size),
                   int(*envSave)(void),
                   int(*readUuid)(unsigned long long *u64Uuid));
  

• Dependency

  • Header File: AudioSeProcess.h

  • Library File: libSE_LINUX.so/ libSE_LINUX.a

---

3. SE DATA TYPE

3.1. Definition of Data Type Related to SE Module

Data Type	Definition
AudioSeInit_t	The structure type of SE algorithm initialization data.
AudioSeNoiseType_e	The enum type of SE algorithm noise selection.
AudioSeConfig_t	The structure type of SE algorithm parameter configuration.
SE_HANDLE	The handle type of SE algorithm.

3.2. AudioSeInit_t

• Description

  Define the configuration parameter structure of SE algorithm.

• Definition

  typedef struct{
  
      int sampleRate;
  
      int bitWidth;
  
      int channel;
  
  }AudioSeInit_t;
  

• Members

  Member Name	Description
  sampleRate	The sampling rate of speech.
  bitWidth	The bitwidth of speech sampling.
  channel	Number of speech channels. Value range [1, 2]. "1" means single channel, while "2" means dual channels.

• Note

  • The sampling bitwidth only supports 16bit, and the sampling rate only supports 16kHz.

  • Dual-channel data should be stored interleaved according to the left and right channels, and the data format should be: L, R, L, R, L, R... and so on.

• Related Data Type and Interface

  IaaSe_Init

3.3. AudioSeNoiseType_e

• Description

  Define the type of noise to be suppressed by SE algorithm.

• Definition

  typedef enum{
  
      IAA_SE_OFFICE_NOISE = 0,
  
      IAA_SE_TRAFFIC_NOISE,
  
      IAA_SE_ALL_NOISE,
  
  }AudioSeNoiseType_e;
  

• Members

  Member Name	Description
  IAA_SE_OFFICE_NOISE	Indoor setting.
  IAA_SE_TRAFFIC_NOISE	Outdoor setting.
  IAA_SE_ALL_NOISE	General setting.

• Note

  • Only IAA_SE_ALL_NOISE is supported currently.

• Related Data Type and Interface

  IaaSe_SetConfig

3.4. AudioSeConfig_t

• Description

  Define the parameter structure of SE algorithm.

• Definition

  typedef struct{
      AudioSeNoiseType_e noiseType;
      int intensity;
      int normalize;
      int smooth;
      int normalizeMode;
      int normalizeVadThreshold;
      int normalizePosition;
  }AudioSeConfig_t;
  

• Members

  Member Name	Description
  noiseType	The noise type to be suppressed.
  intensity	The intensity of noise suppression. Value range: [1~10], and the step length is 1. Setting the value to 1 represents the weakest suppression intensity, while setting the value to 10 represents the greatest suppression intensity. The recommended value is 5.
  normalize	Normalized parameter, which should be enabled when sound volume is small. The recommended value is 20000. When not in use, setting it to 0.
  smooth	Smooth factor when Normalize is enabled. Value range: [1~10]. The larger the value, the smoother it will be. The recommended value is 5.
  normalizeMode	Normalized mode, 0: fixed gain normalize, 1: adaptive normalize
  normalizeVadThreshold	Normalized VAD threshold. Value range: [-80, 0]
  normalizePosition	Normalized position, 0: normalize before speech enhancement, 1: normalize after speech enhancement

• Related Data Type and Interface

  IaaSe_SetConfig

3.5. SE_HANDLE

• Description

  Define the handle type of SE algorithm.

• Definition

  typedef void* SE_HANDLE;
  

• Members

  Member Name	Description

• Note

  N/A.

• Related Data Type and Interface

  IaaSe_Init

  IaaSe_SetConfig

  IaaSe_GetConfig

  IaaSe_Free

  IaaSe_Run

---

4. Error Codes

SE API Error Codes are shown in the following table:

Error Code	Macro Definition	Description
0x00000000	ALGO_SE_RET_SUCCESS	SE algorithm runs successfully.
0x70000401	ALGO_SE_RET_INVALID_LICENSE	Invalid license/Trial period is over.
0x70000402	ALGO_SE_RET_INVALID_HANDLE	Invalid handle.
0x70000403	ALGO_SE_RET_INVALID_SAMPLERATE	Invalid sampling rate.
0x70000404	ALGO_SE_RET_INVALID_BITWIDTH	Invalid sampling bitwidth.
0x70000405	ALGO_SE_RET_INVALID_CHANNEL	Invalid channel number.
0x70000406	ALGO_SE_RET_INVALID_INTENSTIY	Invalid noise suppression intensity.
0x70000407	ALGO_SE_RET_INVALID_NOISETYPE	Invalid noise type.
0x70000408	ALGO_SE_RET_INVALID_NORMALIZE	Invalid normalized amplitude.
0x70000409	ALGO_SE_RET_INVALID_SMOOTH	Invalid smooth factor.
0x7000040A	ALGO_SE_RET_INVALID_NORMALIZE_POS	Invalid normalized position
0x7000040B	ALGO_SE_RET_INVALID_NORMALIZE_MODE	Invalid normalized mode
0x7000040C	ALGO_SE_RET_INVALID_NORMALIZE_VADTHR	Invalid normalized VAD threshold