FreeCalypso > hg > leo2moko-debug
diff chipsetsw/services/Audio/audio_sr_reco.c @ 0:509db1a7b7b8
initial import: leo2moko-r1
author | Space Falcon <falcon@ivan.Harhan.ORG> |
---|---|
date | Mon, 01 Jun 2015 03:24:05 +0000 |
parents | |
children |
line wrap: on
line diff
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/chipsetsw/services/Audio/audio_sr_reco.c Mon Jun 01 03:24:05 2015 +0000 @@ -0,0 +1,478 @@ +/****************************************************************************/ +/* */ +/* File Name: audio_sr_reco.c */ +/* */ +/* Purpose: This file contains all the functions used to manage the */ +/* recognition of a word for the speech recognition module. */ +/* */ +/* Version 0.1 */ +/* */ +/* Date Modification */ +/* ------------------------------------ */ +/* 09 Oct. 2001 Create */ +/* */ +/* Author */ +/* Francois Mazard - Stephanie Gerthoux */ +/* */ +/* (C) Copyright 2001 by Texas Instruments Incorporated, All Rights Reserved*/ +/****************************************************************************/ + +#include "rv/rv_defined_swe.h" +#ifdef RVM_AUDIO_MAIN_SWE + #ifndef _WINDOWS + #include "config/swconfig.cfg" + #include "config/sys.cfg" + #include "config/chipset.cfg" + #endif + + /* include the usefull L1 header */ + #include "l1_confg.h" + + #if (SPEECH_RECO) + #include "rv/rv_general.h" + #include "rvm/rvm_gen.h" + #include "audio/audio_features_i.h" + #include "audio/audio_ffs_i.h" + #include "audio/audio_api.h" + #include "audio/audio_structs_i.h" + #include "audio/audio_var_i.h" + #include "audio/audio_messages_i.h" + #include "rvf/rvf_target.h" + #include "audio/audio_const_i.h" + #include "audio/audio_error_hdlr_i.h" + #include "audio/audio_macro_i.h" + + /* include the usefull L1 header */ + #define BOOL_FLAG + #define CHAR_FLAG + #include "l1_types.h" + #include "l1audio_cust.h" + #include "l1audio_msgty.h" + #include "l1audio_signa.h" + #include "l1audio_cust.h" + + #include <string.h> + + /********************************************************************************/ + /* */ + /* Function Name: audio_sr_reco_convert_parameter */ + /* */ + /* Purpose: Convert the speech reco parameters from the entity to */ + /* the l1 parameters */ + /* */ + /* Input Parameters: */ + /* entity speech reco message */ + /* */ + /* Output Parameters: */ + /* layer 1 speech reco message */ + /* */ + /* Note: */ + /* None. */ + /* */ + /* Revision History: */ + /* None. */ + /* */ + /********************************************************************************/ + void audio_sr_reco_convert_parameter( T_AUDIO_SR_RECO_START *entity_parameter, + T_MMI_SR_RECO_REQ *l1_parameter) + { + l1_parameter->database_id = 0; + l1_parameter->vocabulary_size = entity_parameter->vocabulary_size; + } + + /********************************************************************************/ + /* */ + /* Function Name: audio_sr_reco_send_status */ + /* */ + /* Purpose: This function sends the speech reco status to the */ + /* entity. */ + /* */ + /* Input Parameters: */ + /* status, */ + /* return path */ + /* */ + /* Output Parameters: */ + /* None. */ + /* */ + /* Note: */ + /* None. */ + /* */ + /* Revision History: */ + /* None. */ + /* */ + /********************************************************************************/ + void audio_sr_reco_send_status (T_AUDIO_RET status, UINT8 word_index, T_RV_RETURN return_path) + { + void *p_send_message = NULL; + T_RVF_MB_STATUS mb_status = RVF_RED; + #ifndef _WINDOWS + UINT8 i; + char *p_database; + char *p_word_name; + UINT8 model_name_lengh; + #endif + + while (mb_status == RVF_RED) + { + /* allocate the message buffer */ + mb_status = rvf_get_buf (p_audio_gbl_var->mb_external, + sizeof (T_AUDIO_SR_RECO_STATUS), + (T_RVF_BUFFER **) (&p_send_message)); + + /* If insufficient resources, then report a memory error and abort. */ + /* and wait until more ressource is given */ + if (mb_status == RVF_RED) + { + audio_sr_error_trace(AUDIO_ENTITY_NO_MEMORY); + rvf_delay(RVF_MS_TO_TICKS(1000)); + } + } + + /*fill the header of the message */ + ((T_AUDIO_SR_RECO_STATUS *)(p_send_message))->os_hdr.msg_id = AUDIO_SR_RECO_STATUS_MSG; + + if (status > 0) + { + status = (INT8)((-2) - status); + } + + if (status == SC_NO_ERROR) + /* no error occured dring the recognition */ + { + #ifndef _WINDOWS + /* search the name of the recognized word */ + p_database = p_audio_gbl_var->speech_reco.sr_reco.p_database; + for(i=0; i< word_index; i++) + { + p_database += AUDIO_PATH_NAME_MAX_SIZE; + } + + /* the _sr need to be suppress */ + model_name_lengh = (strlen(p_database) - 3); + + /* copy this path in the message */ + strncpy(((T_AUDIO_SR_RECO_STATUS *)(p_send_message))->word_recognized, + p_database, + model_name_lengh); + p_word_name = (char *)(((T_AUDIO_SR_RECO_STATUS *)(p_send_message))->word_recognized); + p_word_name += model_name_lengh; + *p_word_name = 0; + #endif + } + + /* fill the status parameters */ + ((T_AUDIO_SR_RECO_STATUS *)(p_send_message))->status = status; + + if (return_path.callback_func == NULL) + { + /* send the message to the entity */ + rvf_send_msg (return_path.addr_id, + p_send_message); + } + else + { + /* call the callback function */ + (*return_path.callback_func)((void *)(p_send_message)); + rvf_free_buf((T_RVF_BUFFER *)p_send_message); + } + } + + /********************************************************************************/ + /* */ + /* Function Name: audio_sr_reco_manager */ + /* */ + /* Purpose: This function is called to manage a speech reco */ + /* process. */ + /* */ + /* Input Parameters: */ + /* Audio Speech reco reco Parameters, */ + /* */ + /* Output Parameters: */ + /* None. */ + /* */ + /* Note: */ + /* None. */ + /* */ + /* Revision History: */ + /* None. */ + /* */ + /********************************************************************************/ + void audio_sr_reco_manager (T_RV_HDR *p_message) + { + /* Declare local variables. */ + void *p_send_message; + + /**************** audio_sr_reco_manager function begins *********************/ + switch(p_audio_gbl_var->speech_reco.sr_reco.state) + { + case AUDIO_SR_RECO_IDLE: + { + switch(p_message->msg_id) + { + case AUDIO_SR_RECO_START_REQ: + { + /* save the task id of the entity */ + p_audio_gbl_var->speech_reco.sr_reco.task_id = + ((T_AUDIO_SR_RECO_START*)(p_message))->os_hdr.src_addr_id; + + /* save the return path */ + p_audio_gbl_var->speech_reco.sr_reco.return_path.callback_func = + ((T_AUDIO_SR_RECO_START*)(p_message))->return_path.callback_func; + p_audio_gbl_var->speech_reco.sr_reco.return_path.addr_id = + ((T_AUDIO_SR_RECO_START*)(p_message))->return_path.addr_id; + + /* save the message parameters */ + p_audio_gbl_var->speech_reco.sr_reco.vocabulary_size = + ((T_AUDIO_SR_RECO_START*)(p_message))->vocabulary_size; + p_audio_gbl_var->speech_reco.sr_reco.p_database = + ((T_AUDIO_SR_RECO_START*)(p_message))->p_database; + + /* allocate the buffer for the message to the L1 */ + p_send_message = audio_allocate_l1_message(sizeof(T_MMI_SR_RECO_REQ)); + if ( p_send_message != NULL) + { + /* Convert the entity parameters to the audio L1 parameters */ + audio_sr_reco_convert_parameter((T_AUDIO_SR_RECO_START *)p_message, + (T_MMI_SR_RECO_REQ *)p_send_message); + + /* send the start command to the audio L1 */ + audio_send_l1_message(MMI_SR_RECO_START_REQ, p_send_message); + } + + /* change to the state AUDIO_SR_RECO_WAIT_START_CONFIRMATION */ + p_audio_gbl_var->speech_reco.sr_reco.state = AUDIO_SR_RECO_WAIT_START_CON; + break; + } + + case AUDIO_SR_RECO_STOP_REQ: + { + /* event error - send an error message*/ + audio_sr_reco_send_status (AUDIO_ERROR, 0, + ((T_AUDIO_SR_RECO_STOP *)(p_message))->return_path); + + audio_sr_error_trace(AUDIO_ERROR_STOP_EVENT); + break; + } + } + break; + } /* case AUDIO_SR_RECO_IDLE */ + + case AUDIO_SR_RECO_WAIT_START_CON: + { + switch(p_message->msg_id) + { + case MMI_SR_RECO_START_CON: + { + /* change to the state AUDIO_SR_RECO_WAIT_STOP_COMMAND */ + p_audio_gbl_var->speech_reco.sr_reco.state = AUDIO_SR_RECO_WAIT_STOP_COMMAND; + break; + } + case AUDIO_SR_RECO_STOP_REQ: + /*.Before stopping this task, control that */ + /* stop task id caller = SR reco task id */ + { + if ( p_audio_gbl_var->speech_reco.sr_reco.task_id == + ((T_AUDIO_SR_RECO_STOP*)(p_message))->os_hdr.src_addr_id) + { + /* save the return path */ + p_audio_gbl_var->speech_reco.sr_reco.return_path.callback_func = + ((T_AUDIO_SR_RECO_STOP*)(p_message))->return_path.callback_func; + p_audio_gbl_var->speech_reco.sr_reco.return_path.addr_id = + ((T_AUDIO_SR_RECO_STOP*)(p_message))->return_path.addr_id; + + /* change to the state AUDIO_SR_RECO_WAIT_STOP_CONFIRMATION */ + p_audio_gbl_var->speech_reco.sr_reco.state = AUDIO_SR_RECO_WAIT_START_CON_TO_STOP; + } + else + { + /* A stop request from an other task is sent during a start connection */ + /* event error - send an error message */ + audio_sr_reco_send_status (AUDIO_ERROR, 0, + ((T_AUDIO_SR_RECO_STOP *)(p_message))->return_path); + + audio_sr_error_trace(AUDIO_ERROR_STOP_EVENT); + } + break; + } + case AUDIO_SR_RECO_START_REQ: + { + /* deallocate the vocabulary data buffer */ + rvf_free_buf( + (T_RVF_BUFFER *)(((T_AUDIO_SR_RECO_START *)(p_message))->p_database)); + + /* event error - send an error message*/ + audio_sr_reco_send_status (AUDIO_ERROR, 0, + ((T_AUDIO_SR_RECO_START *)(p_message))->return_path); + + audio_sr_error_trace(AUDIO_ERROR_START_EVENT); + break; + } + } + break; + } /* case AUDIO_SR_RECO_WAIT_START_CON */ + + case AUDIO_SR_RECO_WAIT_START_CON_TO_STOP: + { + switch(p_message->msg_id) + { + case MMI_SR_RECO_START_CON: + { + /* send the stop command to the audio L1 */ + /* allocate the buffer for the message to the L1 */ + p_send_message = audio_allocate_l1_message(0); + if ( p_send_message != NULL) + { + /* send the start command to the audio L1 */ + audio_send_l1_message(MMI_SR_RECO_STOP_REQ, p_send_message); + } + + /* change to the state AUDIO_SR_RECO_WAIT_STOP_CON */ + p_audio_gbl_var->speech_reco.sr_reco.state = AUDIO_SR_RECO_WAIT_STOP_CON; + break; + } + case AUDIO_SR_RECO_START_REQ: + { + /* deallocate the vocabulary data buffer */ + rvf_free_buf( + (T_RVF_BUFFER *)(((T_AUDIO_SR_RECO_START *)(p_message))->p_database)); + + /* event error - send an error message*/ + audio_sr_reco_send_status (AUDIO_ERROR, 0, + ((T_AUDIO_SR_RECO_START *)(p_message))->return_path); + + audio_sr_error_trace(AUDIO_ERROR_START_EVENT); + break; + } + case AUDIO_SR_RECO_STOP_REQ: + { + /* event error - send an error message*/ + audio_sr_reco_send_status (AUDIO_ERROR, 0, + ((T_AUDIO_SR_RECO_STOP *)(p_message))->return_path); + + audio_sr_error_trace(AUDIO_ERROR_STOP_EVENT); + break; + } + } + break; + } + + case AUDIO_SR_RECO_WAIT_STOP_COMMAND: + { + switch(p_message->msg_id) + { + case MMI_SR_RECO_STOP_CON: + { + /* deallocate the vocabulary data buffer */ + rvf_free_buf( + (T_RVF_BUFFER *)(p_audio_gbl_var->speech_reco.sr_reco.p_database)); + + audio_sr_reco_send_status (((T_MMI_SR_RECO_STOP_CON *)p_message)->error_id, + (UINT8)((T_MMI_SR_RECO_STOP_CON *)p_message)->best_word_index, + p_audio_gbl_var->speech_reco.sr_reco.return_path); + + /* change to the state AUDIO_SR_RECO_IDLE */ + p_audio_gbl_var->speech_reco.sr_reco.state = AUDIO_SR_RECO_IDLE; + break; + } + + case AUDIO_SR_RECO_STOP_REQ: + { + /*.Before stopping this task, control that */ + /* stop task id caller = SR reco task id */ + if ( p_audio_gbl_var->speech_reco.sr_reco.task_id == + ((T_AUDIO_SR_RECO_STOP*)(p_message))->os_hdr.src_addr_id) + { + /* save the return path */ + p_audio_gbl_var->speech_reco.sr_reco.return_path.callback_func = + ((T_AUDIO_SR_RECO_STOP*)(p_message))->return_path.callback_func; + p_audio_gbl_var->speech_reco.sr_reco.return_path.addr_id = + ((T_AUDIO_SR_RECO_STOP*)(p_message))->return_path.addr_id; + + /* send the stop command to the audio L1 */ + /* allocate the buffer for the message to the L1 */ + p_send_message = audio_allocate_l1_message(0); + if ( p_send_message != NULL) + { + /* send the start command to the audio L1 */ + audio_send_l1_message(MMI_SR_RECO_STOP_REQ, p_send_message); + } + + /* change to the state AUDIO_SR_RECO_WAIT_STOP_CONFIRMATION */ + p_audio_gbl_var->speech_reco.sr_reco.state = AUDIO_SR_RECO_WAIT_STOP_CON; + } + else + { + /* A stop request from an other task is sent during a start connection */ + /* event error - send an error message */ + audio_sr_reco_send_status (AUDIO_ERROR, 0, + ((T_AUDIO_SR_RECO_STOP *)(p_message))->return_path); + + audio_sr_error_trace(AUDIO_ERROR_STOP_EVENT); + } + break; + } + + case AUDIO_SR_RECO_START_REQ: + { + /* deallocate the vocabulary data buffer */ + rvf_free_buf( + (T_RVF_BUFFER *)(((T_AUDIO_SR_RECO_START *)(p_message))->p_database)); + + /* event error - send an error message*/ + audio_sr_reco_send_status (AUDIO_ERROR, 0, + ((T_AUDIO_SR_RECO_START *)(p_message))->return_path); + + audio_sr_error_trace(AUDIO_ERROR_START_EVENT); + break; + } + } + break; + } /* case AUDIO_SR_RECO_WAIT_STOP_COMMAND */ + + case AUDIO_SR_RECO_WAIT_STOP_CON: + { + switch(p_message->msg_id) + { + case MMI_SR_RECO_STOP_CON: + { + /* deallocate the vocabulary data buffer */ + rvf_free_buf( + (T_RVF_BUFFER *)(p_audio_gbl_var->speech_reco.sr_reco.p_database)); + + audio_sr_reco_send_status (((T_MMI_SR_RECO_STOP_CON *)p_message)->error_id, + (UINT8)((T_MMI_SR_RECO_STOP_CON *)p_message)->best_word_index, + p_audio_gbl_var->speech_reco.sr_reco.return_path); + + /* change to the state AUDIO_SR_RECO_IDLE */ + p_audio_gbl_var->speech_reco.sr_reco.state = AUDIO_SR_RECO_IDLE; + break; + } + case AUDIO_SR_RECO_STOP_REQ: + { + /* event error - send an error message*/ + audio_sr_reco_send_status (AUDIO_ERROR, 0, + ((T_AUDIO_SR_RECO_STOP *)(p_message))->return_path); + + audio_sr_error_trace(AUDIO_ERROR_STOP_EVENT); + break; + } + case AUDIO_SR_RECO_START_REQ: + { + /* deallocate the vocabulary data buffer */ + rvf_free_buf( + (T_RVF_BUFFER *)(((T_AUDIO_SR_RECO_START *)(p_message))->p_database)); + + /* event error - send an error message*/ + audio_sr_reco_send_status (AUDIO_ERROR, 0, + ((T_AUDIO_SR_RECO_START *)(p_message))->return_path); + + audio_sr_error_trace(AUDIO_ERROR_START_EVENT); + break; + } + } + break; + } /* case AUDIO_SR_RECO_WAIT_STOP_CON */ + } /* switch(p_audio_gbl_var->speech_reco.state) */ + } /*********************** End of audio_sr_reco_manager function **********************/ + #endif /* #if (SPEECH_RECO) */ +#endif /* #ifdef RVM_AUDIO_MAIN_SWE */